Pharmacogenomics has demonstrated clinical relevance for selected drug–gene pairs involving genes such as CYP2D6, CYP2C19, CYP2C9, DPYD, SLCO1B1 and NAT2. Genetic variation can influence drug disposition, metabolism, efficacy and the risk of adverse reactions, leading organizations including the Clinical Pharmacogenetics Implementation Consortium and the Dutch Pharmacogenetics Working Group to develop prescribing recommendations. Resources such as PharmGKB, PharmVar, ClinVar and ClinGen support evidence curation, variant interpretation and standardized nomenclature. Implementation nevertheless remains uneven, and even when a guideline exists, inherited genotype does not represent every influence on a patient’s response to a medicine.

The central problem is that enzyme activity is not necessarily fixed by genotype. Concomitant medicines, inflammation, liver dysfunction, aging, pregnancy, smoking and other physiological or environmental conditions can modify metabolic pathways. Strong inhibitors of CYP2D6, including paroxetine, fluoxetine, bupropion and quinidine, can reduce activity to levels comparable to those of poor metabolizers. CYP2C19 and CYP3A activity may also change in response to medications and clinical conditions. Inflammation-mediated suppression of CYP expression is particularly relevant in infectious disease, autoimmune disorders, cancer and critical illness. As a result, a genetically classified normal metabolizer may currently resemble an intermediate or poor metabolizer, while induction can produce the opposite change.

Phenoconversion is especially consequential for medicines with narrow therapeutic indices or those requiring metabolic activation, including tamoxifen, codeine, tramadol and numerous psychotropic agents. Inhibition can increase exposure and produce unexpected toxicity, whereas induction can reduce exposure and contribute to treatment failure. Large cohort evidence cited by the review indicates that concomitant medication use and other non-genetic factors substantially increase the proportion of patients with clinically actionable phenotypes compared with genotype-only prediction. These findings support a dynamic view of metabolic capacity, but they do not mean that every apparent genotype–phenotype discrepancy qualifies as clinically meaningful phenoconversion.

The review locates the potential value of LLMs in a clinical knowledge-integration problem. Medication records, laboratory measurements, diagnostic codes, imaging reports and unstructured notes can all contain modifiers of drug metabolism. Information such as over-the-counter drug use, supplements, smoking cessation, an inflammatory flare or a temporary physiological change may appear only in narrative text. Traditional rule-based systems can struggle when evidence is embedded in free text or must be aligned with a particular treatment period. An LLM might extract these signals, connect them with the patient’s pharmacogenomic report and present them in a form that allows a clinician to evaluate whether the predicted phenotype may no longer reflect current function.

Temporal reasoning would be central to that task. Finding an inhibitor anywhere in a longitudinal record does not establish that it affected metabolism when a medicine was prescribed. Similarly, an infectious or inflammatory condition may coincide with a change in exposure, adverse effects or clinical response without proving that it caused the change. An LLM-assisted workflow could reconstruct when exposures began and ended, distinguish current from historical factors, and relate them to treatment events. The resulting output would identify potential phenoconversion for review rather than confirm it, which is a critical distinction because functional status can be checked through independent measures such as probe-drug testing, therapeutic drug monitoring or a validated biomarker.

The authors propose that LLM outputs could also become inputs to dynamic phenotype models that update metabolic estimates as a patient’s circumstances change. Conventional pharmacogenomic reports generally assign discrete categories such as poor, intermediate, normal or ultrarapid metabolizer status. Dynamic approaches instead combine genotype with current medicines, inflammatory status, organ function, lifestyle and other covariates to estimate the probability that a patient presently falls within a particular metabolic range. These proposals overlap with model-informed precision dosing, population pharmacokinetics, Bayesian forecasting and physiologically based pharmacokinetic modeling. The defensible division of labor is for LLMs to retrieve and structure clinical inputs, while validated pharmacokinetic and pharmacodynamic models estimate exposure, dose requirements and response.

Several scientific and technical barriers remain. LLMs can misinterpret records, generate unsupported statements or make errors in temporal reasoning, and isolated errors may carry clinical consequences. Missing or outdated medication lists, incomplete lifestyle histories and contradictory records can reduce reliability regardless of model sophistication. The magnitude and duration of many non-genetic modifiers are also insufficiently characterized, and routine care lacks a universally accepted reference standard for functional metabolic activity. A system can identify a suspected modifier but cannot derive its effect size, explain an unknown interaction or create a validated dose adjustment when the underlying pharmacological relationship has not been established.

Safe deployment would require validation with real-world clinical data, external testing across institutions and populations, explicit uncertainty estimates, documentation of contributing factors and continuous human oversight. Generalization is a particular concern because electronic record structures, prescribing patterns, language and coding practices vary between health systems, while allele frequencies and genotype–phenotype relationships may differ among ancestry groups. Privacy, cybersecurity, computational cost, model updating and regulatory accountability also complicate implementation. Any system would need to expose supporting quotations or citations and distinguish established evidence from hypotheses, especially when it surfaces unfamiliar combinations of drug–drug–gene or drug–gene–environment interactions.

Future research would need to quantify when changes in metabolic activity meaningfully alter drug exposure, efficacy or toxicity, then test whether LLM-assisted workflows improve medication safety, therapeutic effectiveness, adverse-event detection, healthcare use or patient outcomes. Such systems could integrate with established pharmacogenomic and precision-dosing infrastructure and, over time, analyze large longitudinal datasets to generate hypotheses about previously unrecognized interactions. The perspective therefore presents LLM-enabled phenoconversion detection as a promising but unproven extension of precision pharmacotherapy. Its eventual usefulness will depend not on language-model capability alone, but on pharmacological evidence, complete clinical data, transparent quantitative models and continued oversight by clinicians.

Subject of Research: Potential use of large language models for detecting pharmacogenomic phenoconversion in clinical records

Article Title: The potential role of large language models in addressing phenoconversion in pharmacogenomics

Article References: García-Agúndez, A., & Agúndez, J. A. G. (2026). The potential role of large language models in addressing phenoconversion in pharmacogenomics. Frontiers in Genetics, 17, Article 1987469. https://doi.org/10.3389/fgene.2026.1987469

Image Credits: AI Generated

DOI: 10.3389/fgene.2026.1987469

Keywords: pharmacogenomics, phenoconversion, large language models, precision dosing, drug metabolism, electronic health records, drug interactions, cytochrome P450, clinical decision support, precision medicine, potential, role

The team, led by Luping Zhao, Yan Shen, Yuanjin Zhang, and Xin Wang, evaluated juglone against six major CYP activities in both pooled human liver microsomes and microsomes prepared from rat liver. Using a cocktail of probe substrates—phenacetin for CYP1A2, bupropion for CYP2B6, tolbutamide for CYP2C9, dextromethorphan for CYP2D6, chlorzoxazone for CYP2E1, and midazolam for CYP3A4—they quantified metabolite formation by liquid chromatography–tandem mass spectrometry after incubating the microsomes with juglone concentrations ranging from 1 to 100 micromolar. The screening revealed a clear hierarchy of susceptibility in the human preparations, with CYP2C9 and CYP2E1 standing out well ahead of the rest.

In human liver microsomes, CYP2C9 was the most vulnerable activity, with a half-maximal inhibitory concentration of 8.12 micromolar. Kinetic experiments that varied both substrate and inhibitor concentrations pointed to competitive inhibition, with an apparent inhibition constant of 4.97 micromolar—the lowest Ki recorded among the human activities examined. CYP2E1 was the next most susceptible, showing an IC50 of 12.67 micromolar and a mixed inhibition pattern, with apparent Ki and alpha-Ki values of 19.59 and 8.74 micromolar respectively. By contrast, CYP1A2 and CYP2B6 were only weakly inhibited, with IC50 values near 50 micromolar, and CYP2D6 and CYP3A4 were not inhibited by half even at the highest concentration tested.

The rat microsomes told a different story. There, CYP1A2 was the most strongly inhibited activity, with an IC50 of 12.03 micromolar, followed by CYP2E1 at 18.50 micromolar. Follow-up kinetic analysis identified mixed inhibition of rat CYP1A2, with an apparent Ki of 12.31 micromolar, and competitive inhibition of rat CYP2E1, with an apparent Ki of 19.89 micromolar. Notably, rat CYP2C11—the nominal counterpart of human CYP2C9—was considerably less sensitive, with an IC50 of 33.39 micromolar. The authors are careful to note that CYP2C11 is not a one-to-one equivalent of human CYP2C9, and that CYP abundance differs between the two species, so the rat profile should be used to design animal studies rather than as a quantitative surrogate for human risk.

One of the more consequential findings concerns what the study did not detect. Time-dependent inhibition experiments, in which microsomes were preincubated with juglone for up to twenty minutes before the probe reaction was started, showed no consistent progressive loss of enzyme activity for any of the selected CYPs. That argues against mechanism-based inactivation under the tested conditions, though the authors acknowledge important caveats: the preincubation window was short, the concentration range limited, and—critically—no dedicated positive control for time-dependent inhibition was included. Definitive conclusions about reversibility, and any measurement of the inactivation parameters kinact and KI, will require follow-up experiments with appropriate controls and longer preincubation periods.

The species comparison around CYP2E1 is particularly instructive. The apparent Ki values in human and rat microsomes were numerically almost identical—19.59 versus 19.89 micromolar—yet the fitted inhibition models differed, with mixed inhibition in the human system and competitive inhibition in the rat system. The authors resist the temptation to read this as a genuine species difference. The two assays used different microsomal protein concentrations and buffer compositions, factors known to alter apparent inhibitory potency and catalytic activity. Without matched conditions and verified reaction linearity, they write, similar Ki values cannot be taken as evidence of equivalent intrinsic susceptibility, and the observed model differences cannot be attributed solely to species.

To add structural context, the team docked juglone into crystal structures of human CYP2C9 and CYP2E1. The docking produced GlideScores of −6.758 and −7.502 kilocalories per mole respectively, and the top-ranked CYP2C9 pose placed the naphthoquinone ring close to the heme catalytic center, flanked by hydrophobic residues including Leu102, Ala103, Leu208, and Val237, with a predicted hydrogen bond between a juglone carbonyl and Arg108. Interestingly, the docking scores did not reproduce the experimental potency ranking—CYP2E1 scored better yet was inhibited more weakly in human microsomes—a discrepancy the authors attribute to the limits of empirical scoring functions, which cannot be translated directly into inhibition constants and do not model substrate competition or the enzyme–substrate–inhibitor states relevant to mixed inhibition.

The translational question—could juglone-containing products actually perturb the metabolism of prescription drugs?—remains open, and the paper is unusually candid about why. Reliable human plasma data for unchanged juglone under relevant use conditions are simply unavailable. Preclinical evidence suggests partial absorption in rats, formulation-dependent half-lives, and more than 99 percent binding to serum albumin, but none of this establishes the unbound systemic concentration that would determine whether the micromolar inhibition seen in vitro is ever approached in vivo. The authors convert their key numbers for perspective: the CYP2C9 IC50 of 8.12 micromolar corresponds to roughly 1.41 micrograms per milliliter, and the Ki of 4.97 micromolar to about 0.866 micrograms per milliliter—nominal in vitro concentrations, not clinical thresholds. Measuring the unbound, unchanged compound in plasma is flagged as the essential next step.

The study also fits into a broader research program. The same laboratory previously showed that the related naphthoquinones plumbagin, shikonin, and deoxyshikonin inhibited all six tested human CYP activities, whereas juglone’s effects concentrated on CYP2C9 and CYP2E1. That contrast underscores a point the authors emphasize repeatedly: structural relatedness alone does not define a shared inhibition profile, and each compound must be assessed on its own. It also matters clinically, because CYP2C9 clears drugs with narrow therapeutic windows, and inhibitors of this enzyme have well-documented potential to raise exposure to anticoagulants and anticonvulsants.

The authors are equally transparent about the study’s statistical and analytical limits. All replicates were technical incubation replicates within a single pooled microsomal preparation, so the reported variability says nothing about inter-individual differences in enzyme activity. The apparent Ki and alpha-Ki values were derived from Lineweaver–Burk plots and secondary replots, now accompanied by 95 percent confidence intervals, and were cross-checked with global nonlinear regression and the corrected Akaike information criterion. For human CYP2C9, the lowest-AICc competitive model agreed with the graphical analysis, yielding a Ki of 5.11 micromolar versus 4.97 micromolar from the replots. For human CYP2E1, the mixed model was only modestly favored over a pure noncompetitive alternative, and some alpha-Ki estimates—particularly for the rat enzymes—were poorly constrained. Comprehensive analytical validation of the LC-MS/MS method under the study’s exact conditions was not performed, leaving quantitative bias incompletely characterized.

Where does this leave juglone? The compound, enriched in Pericarpium Juglandis—the green walnut husk used in Chinese medicine as Qinglongyi—has attracted attention as a scaffold for derivatives targeting cancer, inflammation, and bacterial quorum sensing. This study does not sound an alarm for consumers, but it does something arguably more useful: it converts a vague concern about natural product–drug interactions into a concrete, testable hypothesis centered on CYP2C9. The recommended roadmap includes recombinant human CYP2C9 experiments, measurements of plasma and microsomal protein binding, metabolite identification to check whether juglone’s breakdown products also inhibit CYPs, and in vivo pharmacokinetic studies that quantify unchanged juglone under defined dosing and formulation conditions. Only with that exposure data, combined with binding-adjusted inhibition estimates and knowledge of how much of a co-administered drug’s clearance depends on CYP2C9, can a defensible prediction of interaction risk be made. For now, the message is measured: juglone is a potent CYP2C9 inhibitor in a test tube, and that fact alone makes it a compound worth watching.

Subject of Research: In vitro inhibition of cytochrome P450 enzymes by the natural naphthoquinone juglone in human and rat liver microsomes

Article Title: Inhibition of cytochrome P450 activities by juglone in human and rat liver microsomes: kinetic characterization and implications for natural product–drug interaction assessment

Article References: Zhao, L., Shen, Y., Xu, X., Liu, L., Huang, J., Liang, C., Zhang, Y., & Wang, X. (2026). Inhibition of cytochrome P450 activities by juglone in human and rat liver microsomes: kinetic characterization and implications for natural product–drug interaction assessment. Frontiers in Pharmacology, 17, Article 1968844. https://doi.org/10.3389/fphar.2026.1968844

Image Credits: AI Generated

DOI: 10.3389/fphar.2026.1968844

Keywords: juglone, cytochrome P450, CYP2C9, CYP2E1, liver microsomes, drug metabolism, enzyme kinetics, natural product-drug interactions, naphthoquinone, pharmacokinetics, molecular docking, time-dependent inhibition

The research, led by Hongliang Cui, Jiaxin Li, Yangbo Fu, and Feiteng Sun, with corresponding authors Lei He, Zhiran Li, Fei Sun, and Jun Yu, focuses on the Drosophila testis, one of the most powerful and best-understood model systems for studying adult stem cell biology. Within this organ, germline stem cells sit alongside another population known as cyst stem cells, or CySCs. The CySCs are somatic cells, meaning they do not make sperm themselves. Instead, they envelop the germline cells, provide essential structural and biochemical support, and help orchestrate the step-by-step process by which a stem cell ultimately becomes a mature sperm cell. The balance between keeping CySCs in their stem cell state and allowing them to differentiate is delicate, and the mechanisms that maintain this balance have remained incompletely understood.

The team set out to determine what happens when a single gene in the CySC lineage is manipulated. Through targeted genetic experiments that raised or lowered the levels of Mppa specifically in these somatic support cells, the researchers uncovered two distinct problems. First, the maintenance of the cyst stem cells themselves was disrupted, meaning the population of supporting cells that anchors the niche began to falter. Second, and perhaps more strikingly, the differentiation of the neighboring germ cells was impaired. This second effect is what scientists call a nonautonomous function: Mppa acts in one cell type, the CySCs, but its consequences ripple outward to affect a completely different cell type, the germline.

Nonautonomous regulation is a recurring theme in stem cell biology, but the specific mechanism uncovered here adds an unexpected twist. When the researchers dug into what Mppa actually does at the molecular level, they found that it is a conserved mitochondrial protein with a central role in maintaining ATP homeostasis. ATP, or adenosine triphosphate, is the universal energy currency of the cell, and mitochondria are the organelles that generate most of it. In other words, the failure of germ cells to differentiate properly in these experiments was not caused by a signaling molecule or a transcription factor acting directly on the germline. It was caused, at least in part, by an energy supply problem in the supporting cells.

To understand how a mitochondrial defect in CySCs could translate into a developmental failure, the team turned to transcriptomics, the systematic measurement of gene expression across the genome. The transcriptomic analysis revealed that Mppa’s influence converged on mitochondrial function, with particular involvement of pathways related to cytochrome P450 enzymes. Cytochrome P450 proteins are best known for their roles in metabolism and detoxification, but in this context they emerged as part of the regulatory landscape through which Mppa shapes mitochondrial performance. The finding suggests that the energy state of the niche cells is monitored and managed through a coordinated transcriptional program, and that Mppa sits near the top of that program.

The researchers did not stop at gene expression. To identify the physical partners of Mppa, they performed a proteomic screen using liquid chromatography coupled with tandem mass spectrometry, or LC–MS/MS, a technique that can catalog the proteins that associate with a molecule of interest inside cells. The results painted a picture of Mppa as deeply embedded in the cell’s energy machinery. Its binding partners were enriched for proteins involved in diverse energy metabolism processes, oxidation–reduction reactions, and the maintenance of cellular redox balance, the equilibrium between oxidizing and reducing agents that cells must constantly manage to avoid damage.

Among the many associations detected in the mass spectrometry screen, the team went on to validate specific interactions experimentally. One confirmed binding partner was UQCR-C1, a component of the ubiquinol–cytochrome c reductase complex, better known as Complex III of the mitochondrial electron transport chain. This complex is a critical relay in the series of molecular steps that transfer electrons derived from food and ultimately drive the production of ATP. The physical association between Mppa and a core component of the respiratory chain provides a concrete biochemical link between the protein and the energy homeostasis it helps to regulate, grounding the genetic and transcriptomic observations in a defined molecular interaction.

Taken together, the study proposes a model in which Mppa functions as a kind of metabolic custodian for the testicular stem cell niche. By sustaining ATP levels and supporting mitochondrial and redox homeostasis in CySCs, Mppa enables those cells to remain stem cells and to perform their supportive duties toward the germline. When Mppa is lost or reduced, the energy balance of the niche falters, the CySC population becomes unstable, and the germ cells that depend on them fail to progress through their differentiation program. The germline defect, in this view, is a downstream consequence of a metabolic failure in a neighboring cell type, a striking illustration of how deeply interdependent the cells of a tissue niche really are.

The choice of the Drosophila testis as the experimental system is significant. The fly testis has long served as a paradigm for studying how stem cells are maintained and how niche signals control fate decisions, and many principles discovered there have proven relevant to stem cell biology across species. Because the germline and its somatic support cells can be manipulated independently with genetic tools, the system is uniquely suited to dissecting nonautonomous interactions of exactly the kind described here. The fact that Mppa is described as a conserved mitochondrial protein raises the possibility that similar energy-dependent mechanisms operate in other stem cell niches, including those of mammals.

The clinical resonance of the work lies in male fertility. Spermatogenesis disorders linked to disrupted germ cell differentiation are a recognized cause of male infertility, and the mechanisms behind many such disorders remain poorly defined. By showing that the energy equilibrium of somatic niche cells can govern germ cell differentiation nonautonomously, the study opens a new conceptual avenue for understanding and potentially investigating infertility: rather than looking only at the germ cells themselves, researchers and clinicians may need to consider the metabolic health of the supporting cells that surround them. The work was supported by the National Key Research and Development Program of China, the National Natural Science Foundation of China, the Nantong Natural Science Foundation, and the Qinglan Project of Jiangsu Province, and all experiments were conducted on Drosophila melanogaster, requiring no ethical approval for human or animal subjects. As with any fly genetics study, translating these findings to human biology will require further work, but the study offers a compelling demonstration that the road to healthy sperm runs through the mitochondria of the cells next door.

Subject of Research: Nonautonomous mitochondrial regulation of cyst stem cell maintenance and germ cell differentiation in the Drosophila testis

Article Title: Mppa sustains cyst stem cell maintenance and orchestrates germ cell differentiation through nonautonomous regulation of mitochondrial homeostasis in Drosophila testes

Article References: Cui, H., Li, J., Fu, Y., Sun, F., Wang, X., Chen, X., Wu, J., Chu, Y., Wang, C., Dong, J., He, L., Li, Z., Sun, F., & Yu, J. (2026). Mppa sustains cyst stem cell maintenance and orchestrates germ cell differentiation through nonautonomous regulation of mitochondrial homeostasis in Drosophila testes. Cell Death Discovery. https://doi.org/10.1038/s41420-026-03397-0

Image Credits: AI Generated

DOI: 10.1038/s41420-026-03397-0

Keywords: Drosophila, spermatogenesis, cyst stem cells, stem cell niche, mitochondria, ATP homeostasis, germ cell differentiation, cytochrome P450, UQCR-C1, redox balance, male fertility, Cell Death Discovery

The study, led by co-first authors Xia Xiao and Lin Lin, with corresponding authors Fumei Xin, Jiming Liu and Jarkko Salojärvi, took advantage of the shrub’s own biology. Because different tissues specialize in different metabolic tasks, the researchers profiled roots, stems, leaves and seeds, combining gene expression data with widely targeted metabolomic measurements. This dual approach allowed them to characterize tissue-specific specialized metabolism across the whole plant rather than inferring alkaloid chemistry from a single organ. Metabolomic profiling detected a diverse array of secondary metabolites, including several quinolizidine alkaloids whose abundance patterns varied distinctly from tissue to tissue, providing the raw signal needed to trace where and how these compounds accumulate.

Technically, the work relied on co-accumulation and co-expression network analyses, methods that cluster metabolites and genes into modules based on shared behavior across samples. When a group of metabolites rises and falls together across tissues, that co-accumulation pattern suggests they may share a biosynthetic route or regulatory context. Similarly, when genes are co-expressed, they may be controlled by the same transcription factors or respond to the same developmental cues. By identifying metabolite modules associated with quinolizidine alkaloid abundance and gene modules associated with those metabolites, the team narrowed thousands of possibilities down to a manageable set of candidates.

The integrative transcript-metabolite analyses then prioritized specific enzyme families as likely players in the alkaloid pathway. Among the top candidates were genes encoding copper amine oxidases, or CuAO enzymes, which in other legumes help convert diamines such as cadaverine into the aminated precursors that feed quinolizidine alkaloid skeletons. Dihydrodipicolinate synthase, or DHDPS, genes also emerged, pointing toward the lysine-derived biosynthetic branch that supplies cadaverine. Putrescine N-methyltransferase, or PMT, genes were flagged as well, hinting at methylated polyamine intermediates, while cytochrome P450, or CYP, genes appeared repeatedly across the correlation and multivariate analyses, suggesting they may catalyze oxidative tailoring steps that diversify the alkaloid structures.

Several CYP candidates were recovered again and again across different analytical strategies, a pattern that strengthens their candidacy. Because cytochrome P450 enzymes are notoriously versatile, catalyzing hydroxylations, epoxidations and other modifications across plant specialized metabolism, their repeated recovery in this context suggests they may be responsible for the structural variety observed among the alkaloids the team detected. The researchers also found that (-)-sparteine and lupinine shared co-accumulation patterns and overlapping candidate gene sets, indicating that these two alkaloids may be produced through partially shared regulatory or metabolic contexts rather than entirely separate routes.

Beyond the enzyme families themselves, the integrative analyses prioritized candidate transcriptional regulators potentially associated with quinolizidine alkaloid biosynthesis. These regulatory candidates, if validated, could reveal how the shrub switches alkaloid production on or off in response to developmental stage, tissue type or environmental stress. Understanding such switches is crucial not only for basic science but also for any future effort to boost alkaloid yields through breeding or metabolic engineering, since transcription factors often control entire biosynthetic programs rather than single steps.

The researchers are careful to frame their findings as a foundation rather than a finished story. The prioritized enzymes, metabolite modules and regulatory candidates establish a starting point for future biochemical and genetic validation, meaning the actual catalytic functions of these genes still need to be confirmed through enzyme assays, gene knockouts or heterologous expression experiments. Correlation, however strong, does not prove causation, and the team’s own language reflects that caution, describing candidates as potentially associated with the pathway rather than definitively responsible for it.

Even so, the study delivers resources that extend well beyond a single species. The transcriptomic and metabolomic datasets now serve as reference points for discovering quinolizidine alkaloid candidate genes in other legumes, a plant family that includes several alkaloid-rich medicinal and forage species. Because the shrub is endemic to the Qinghai-Xizang Plateau and faces pressure from habitat change, having molecular resources for it also supports conservation and cultivation efforts, including the breeding programs already underway through regional forestry initiatives in Xizang.

The work also illustrates a broader shift in how plant specialized metabolism is being studied. Rather than sequencing a genome and hoping to spot alkaloid genes by similarity alone, researchers increasingly pair transcriptomics with metabolomics, letting the chemistry itself guide the gene hunt. This chemistry-first strategy is particularly valuable for pathways like quinolizidine alkaloid biosynthesis, where the enzymes involved are often lineage-specific and poorly conserved, making homology-based searches unreliable. By anchoring the analysis in measured metabolite abundance across tissues, the team ensured that their candidate list reflects the plant’s actual metabolic behavior.

For now, the shrub’s chemical secrets remain partially locked, but the candidate genes identified in this study give researchers a concrete map to follow. If future validation confirms the roles of the CuAO, DHDPS, PMT and CYP candidates, and if the prioritized transcription factors turn out to control the pathway, Sophora moorcroftiana could become a model for understanding how a plateau-adapted legume manufactures its defensive alkaloids. That, in turn, could open doors to engineering alkaloid production in related crops, developing new medicinal compounds, and preserving a unique piece of the Qinghai-Xizang Plateau’s botanical heritage.

Subject of Research: Candidate quinolizidine alkaloid biosynthetic genes in Sophora moorcroftiana

Article Title: Integrated transcriptomic and metabolomic analysis prioritizes candidate quinolizidine alkaloid biosynthetic genes in Sophora moorcroftiana

Article References: Xiao, X., Lin, L., Zang, J., Xu, W., Wang, Y., Gesang, Q., Li, B., Da, B., Wang, X., Zhang, C., Xiao, C., Xin, F., Liu, J., & Salojärvi, J. (2026). Integrated transcriptomic and metabolomic analysis prioritizes candidate quinolizidine alkaloid biosynthetic genes in Sophora moorcroftiana. BMC Plant Biology. https://doi.org/10.1186/s12870-026-10161-0

Image Credits: AI Generated

DOI: 10.1186/s12870-026-10161-0

Keywords: Sophora moorcroftiana, quinolizidine alkaloids, biosynthesis, transcriptomics, metabolomics, specialized metabolism, candidate genes, medicinal plant, Qinghai-Xizang Plateau, cytochrome P450, Integrated, transcriptomic

The research team, led by Andrew Disharoon of the Medical University of South Carolina and Aaron J. Masino of Clemson University, assembled a dataset of 377,628 drug-pair samples drawn from DrugBank and the RxNav Drug and Drug Interaction dataset, with chemical structures retrieved through the PubChem API and therapeutic target information from the Therapeutic Target Database. Rather than relying on a small set of broad interaction categories, the team processed the free-text interaction descriptions in these databases—for example, sentences like “the metabolism of (R)-warfarin can be increased when combined with acetaminophen”—and extracted 168 normalized interaction templates by masking out the drug names. Each sample therefore consists of a drug pair, structural and target information for both compounds, and a set of binary labels indicating which of the 168 interaction types apply. Careful preprocessing removed formulation-specific terms such as “hydrochloride” or “maleate” so that, for instance, metoprolol succinate and metoprolol tartrate are treated as the same base drug, and a synonym list unified names like acetaminophen and paracetamol. Interactions with fewer than 50 samples were dropped to ensure enough data for reliable training.

A central design choice distinguishes InfoMedex from earlier approaches: the way molecules are turned into sequences a neural network can read. Instead of treating SMILES strings—compact text encodings of molecular structure—as ordinary character strings, the team converted them to Atom-in-SMILES notation, in which each token explicitly encodes an atom together with its local bonding environment, including ring membership, aromaticity, and attached hydrogens. This chemically meaningful tokenization preserves structural context that character-level splitting or compressed molecular fingerprints can lose, and it is particularly well suited to transformer architectures, where positional embeddings and self-attention can capture relationships between distant substructures. Prior work has shown that Atom-in-SMILES representations better capture local chemical context than character-level tokenization, facilitating substructure-level reasoning.

The architecture itself is a hybrid of transformer and convolutional components processing three input streams. Tokenized Atom-in-SMILES sequences for the drug of interest and for the interacting drug are each converted into dense embeddings of dimension 512 with learned positional encodings. Gene features derived from the Therapeutic Target Database, one-hot encoded across 1,143 features, pass through a two-layer feed-forward network into the same embedding space and are replicated across the sequence length. The three streams are concatenated into a unified sequence of size 3,072 and processed by a transformer encoder with eight attention heads, pre-layer normalization, and padding masks to keep padding tokens out of attention computations. A convolutional stack of four 1D layers with 124 filters and kernel size 7, ReLU activations, a dropout layer, and a linear classification head then produces logits for all 168 interaction labels. The full model contains roughly 70.35 million learnable parameters and was trained with sigmoid focal loss, an objective designed for severe class imbalance, reaching peak validation performance after an average of 22.8 epochs.

The performance results are striking, though they come with an important caveat. Under the primary evaluation, in which drug pairs are split randomly but individual drugs may appear in both training and test sets, the model achieved a micro-averaged area under the precision-recall curve of 0.857 and a macro mean AUPRC of 0.825 across the 168 labels. Accuracy averaged 0.997 and AUC-ROC 0.998, while precision-recall analysis showed the model maintaining precision above 70 percent at recall values as high as 80 percent. Given that only 0.603 percent of the more than 6.3 million test predictions were positive labels, the researchers deliberately relied on precision-recall curves rather than ROC curves, which prior studies have shown to be essentially uninformative under extreme class imbalance. Stratified analysis across frequency quantiles—from interactions positive in as few as 0.013 percent of samples up to the most common at 13.4 percent—showed median precision above 0.8 across all quantiles. For three very rare interaction types the model’s predicted probabilities failed to exceed the 0.5 decision threshold, but the probability distributions for positive and negative samples were largely separated, suggesting appropriate model uncertainty rather than outright failure.

That caveat emerged from a secondary, more demanding evaluation. When the team held out 10 percent of the drugs—86 compounds appearing as the drug of interest in at least one sample—and excluded every sample containing them, mean AUPRC fell from 0.825 to 0.341 across 142 evaluable labels. In other words, InfoMedex generalizes well to unseen drug pairs involving drugs it has encountered elsewhere in training, but predicting interactions for entirely unseen molecules remains substantially harder. This distinction matters enormously for the model’s ultimate use case: flagging risks for newly approved drugs that have no interaction history at all. The authors are candid that this limitation should temper expectations for near-term deployment on novel compounds.

Perhaps the most intriguing contribution is the explainability analysis. Using Integrated Gradients, a gradient-based attribution method implemented in the Captum library, the team computed the contribution of each atom token to selected predictions, aggregating per-sample ranks across many partner drugs to obtain a robust estimate of which molecular regions matter most. Because a given drug’s token sequence and input position are fixed across samples, the averaging marginalizes only over the partner drug, yielding a well-defined estimand: the typical within-drug importance ranking of each token over the empirical distribution of partners. Rank aggregation was chosen over magnitude-based averaging because it is robust to samples carrying disproportionate attribution mass, and interquartile ranges were reported to measure rank stability.

The case studies focused on two well-characterized cytochrome P450 inhibitors. For bupropion, a CYP2D6 inhibitor represented by 29 atom tokens, the highest-importance tokens clustered on the carbonyl-centered motif—the ketone carbonyl carbon had a mean rank of 1.88—and on the bulky N-alkyl tert-butylamine region, including the quaternary carbon at mean rank 6.42 and the basic nitrogen. Both patterns align with known pharmacology: the carbonyl is the starting point for reductive metabolites that show stronger CYP2D6 inhibition than the parent drug, and CYP2D6 substrates and inhibitors frequently present a basic nitrogen that contributes to binding in the enzyme’s tight active site. The meta-chloro phenyl region also ranked prominently, consistent with evidence that this motif discourages aromatic hydroxylation and redirects metabolism toward CYP2B6, which produces hydroxybupropion—the metabolite responsible for roughly two-thirds of bupropion’s in vivo CYP2D6-inhibiting activity.

Ritonavir, the HIV protease inhibitor deployed clinically precisely because of its potent CYP3A4 inhibition, presented a different picture. Represented by 83 atom tokens with 18 points of rotation, the molecule drew the model’s strongest emphasis to the central scaffold, where the hydroxyl token had the highest mean rank of 1.39 with a narrow interquartile range of 1.0 to 1.75 across 18 partner drugs, followed by an adjacent stereocenter. These atoms sit near the benzyl and phenyl bulk that provides steric stabilization, helping hold the drug in a binding-ready configuration within the CYP3A4 cavity. Structural studies have shown ritonavir is a type II ligand that coordinates the heme iron through a thiazole nitrogen while packing hydrophobic groups tightly into the active site, effectively locking the molecule into a high-affinity pose with a slow off-rate. The model also emphasized carbonyl-linked oxygen tokens and the methylene linker connecting the heteroaryl ring to the tertiary amine arm, features that plausibly govern how the bulky aromatic half and the heme-facing heteroaryl half engage the enzyme. Notably, some highly ranked tokens were non-atom structural signposts such as parentheses, which the authors suggest the model uses to guide its structural reading of the molecule rather than as chemically interpretable features.

Taken together, the results position InfoMedex as a scalable, interpretable complement to post-marketing surveillance and formal interaction studies, of which only a limited subset of clinically relevant combinations can feasibly be evaluated under FDA guidance. The ablation finding that randomly perturbing the target features did not significantly change performance metrics suggests the current one-hot gene encodings underexploit mechanistic context, and the authors propose graph neural networks or cross-attention architectures as avenues to better fuse drug-target data. The held-out-drug results temper enthusiasm for novel compounds, and the explainability analysis covered only interactions with well-studied CYP-mediated mechanisms. Still, the demonstration that a transformer reading molecules atom by atom can both predict 168 interaction types with high precision and point to the specific chemical motifs responsible marks a meaningful step toward AI systems that do not just flag drug interactions but can generate mechanistic hypotheses about why they occur—exactly the kind of transparent reasoning that clinical and regulatory adoption will demand.

Subject of Research: Deep learning prediction and explainability of drug-drug interactions using Atom-in-SMILES tokenization and a multimodal CNN-transformer architecture

Article Title: InfoMedex: drug-drug interaction prediction via a multimodal CNN-transformer model

Article References: Disharoon, A., Pasupuleti, S., Thurston, C., Fiore, M., McKenzie, S., Smith, A., & Masino, A. J. (2026). InfoMedex: drug-drug interaction prediction via a multimodal CNN-transformer model. Frontiers in Drug Safety and Regulation, 6, Article 1841035. https://doi.org/10.3389/fdsfr.2026.1841035

Image Credits: AI Generated

DOI: 10.3389/fdsfr.2026.1841035

Keywords: drug-drug interactions, deep learning, transformer, Atom-in-SMILES, explainable AI, Integrated Gradients, cytochrome P450, pharmacovigilance, DrugBank, bupropion, ritonavir, precision medicine

The investigation began with a deceptively simple comparative design. The researchers assembled publicly available transcriptomic datasets from three groups: smokers with lung cancer, smokers without lung cancer, and never-smokers who nevertheless developed the disease. This tripartite structure allowed them to isolate molecular signatures that are specifically tied to the interaction between smoking exposure and cancer development, rather than signatures that merely reflect the presence of a tumor. Differential expression analysis flagged genes whose activity changed across the groups, and a complementary technique called weighted gene co-expression network analysis, or WGCNA, grouped thousands of genes into modules whose coordinated activity correlated with smoking status and disease state.

From this convergence of approaches, the team distilled a set of 16 smoking-specific lung cancer-related genes. Functional enrichment analysis revealed that these genes cluster heavily in pathways devoted to cytochrome P450-mediated xenobiotic metabolism, oxidoreductase activity, and chemical carcinogenesis. In plain terms, the signature is dominated by the cellular machinery that metabolizes foreign chemicals, the very machinery that tobacco smoke confronts with each puff. This finding is biologically coherent: cytochrome P450 enzymes are double-edged swords in toxicology, capable of detoxifying harmful compounds but also of converting procarcinogens in tobacco smoke into reactive metabolites that can damage DNA and initiate the mutational cascade leading to cancer.

To move from correlation toward a more discriminating molecular signature, the researchers turned to machine learning algorithms trained to distinguish lung cancer samples from controls based on gene expression patterns. The analysis converged on a compact four-gene classifier comprising CYP1A1, CYP1B1, HLA-DRA, and IDO1. Within this signature, CYP1A1 and CYP1B1 made substantial contributions to the classification of lung cancer, suggesting that the expression levels of these two enzymes carry considerable diagnostic information. HLA-DRA, a major histocompatibility complex class II gene involved in antigen presentation, and IDO1, an enzyme that catabolizes tryptophan and modulates immune responses, add immune dimensions to the signature, hinting that smoking-related lung carcinogenesis involves both carcinogen metabolism and immunological remodeling.

One of the most consequential steps in the study was the application of Mendelian randomization, a statistical technique that uses naturally occurring genetic variants as instruments to probe whether an exposure-gene relationship is causal rather than merely associative. Drawing on genetic data resources including the GWAS Catalog and the eQTLGen Consortium, the team found that genetically predicted expression levels of CYP1A1 and CYP1B1 were positively associated with lung cancer risk. Because genetic variants are randomly allocated at conception in a manner analogous to a randomized trial, this result strengthens the argument that higher expression of these two enzymes is not simply a downstream consequence of the disease but may actively contribute to carcinogenesis.

Single-cell transcriptomic analysis added a spatial dimension to the story. By examining gene expression at the resolution of individual cells, the researchers discovered that CYP1A1 and CYP1B1 are predominantly expressed in epithelial cells, the cell type that lines the airways and gives rise to the vast majority of lung cancers. This cellular localization is significant because it places the enzymes at the very front line of tobacco smoke exposure, where inhaled carcinogens first contact tissue. Epithelial cells bearing elevated levels of these metabolic enzymes would be positioned to both encounter and bioactivate tobacco carcinogens, a combination that could accelerate the accumulation of oncogenic mutations.

Molecular docking studies provided a structural hypothesis for how these enzymes interact with tobacco-related carcinogens. Docking is a computational method that predicts how small molecules fit into the binding pockets of proteins, and the analysis suggested plausible physical interactions between the CYP1A1 and CYP1B1 proteins and carcinogenic compounds found in tobacco smoke. While docking predictions require experimental confirmation, they offer a mechanistic bridge between the genetic and transcriptomic findings: the enzymes that are overexpressed in smokers with lung cancer are also structurally capable of engaging the very chemicals that smoke delivers to the lung.

The final layer of evidence came from the laboratory. The researchers exposed BEAS-2B cells, an immortalized line of human bronchial epithelial cells widely used to model the airway lining, to cigarette smoke extract. The treatment significantly increased CYP1A1 and CYP1B1 expression at both the messenger RNA and protein levels, demonstrating that the upregulation observed in patient datasets can be reproduced experimentally when airway epithelial cells encounter smoke constituents. This in vitro validation closes an important loop, showing that the association detected in human tissue is at least consistent with a direct response of airway cells to tobacco exposure.

Taken together, the findings sketch a coherent mechanistic model. Tobacco smoke delivers a complex cocktail of carcinogens to the bronchial epithelium, where cells respond by ramping up expression of CYP1A1 and CYP1B1. These enzymes then metabolize smoke constituents, and in doing so may generate reactive intermediates that damage DNA, deplete antioxidant defenses, and drive oxidative stress responses. Over years of repeated exposure, this cycle of activation and damage could push epithelial cells down the path toward malignancy. The positive Mendelian randomization results lend genetic support to this causal interpretation, while the single-cell and docking analyses explain where in the tissue and through which molecular interactions the process is likely to unfold.

The clinical implications are cautiously promising. If CYP1A1 and CYP1B1 expression levels can be reliably measured, for example in airway brushings, sputum, or blood-derived signatures, they could serve as biomarkers for identifying smokers whose lungs show molecular evidence of heightened carcinogenic activation, allowing earlier and more intensive screening. The four-gene signature that also includes HLA-DRA and IDO1 may further refine such risk stratification. At the same time, the authors and observers alike emphasize that these results arise from computational and cell-line analyses of public datasets, and that the enzymes’ driver status must be confirmed in prospective cohorts and functional studies before they can inform clinical practice. Nevertheless, by converging multiple independent lines of evidence on two enzymes that sit at the intersection of tobacco chemistry and lung biology, the study offers one of the clearest molecular accounts to date of why smoking and lung cancer are so tightly, and so variably, linked.

Subject of Research: Molecular mechanisms of smoking-related lung carcinogenesis involving CYP1A1 and CYP1B1

Article Title: Integrated identification of CYP1A1 and CYP1B1 as potential drivers of smoking-related lung carcinogenesis

Article References: Yu, X., Yang, R., Xu, Y., Li, L., & He, Z. (2026). Integrated identification of CYP1A1 and CYP1B1 as potential drivers of smoking-related lung carcinogenesis. 3 Biotech, 16(11), Article 465. https://doi.org/10.1007/s13205-026-05099-x

Image Credits: AI Generated

DOI: 10.1007/s13205-026-05099-x

Keywords: lung cancer, smoking, CYP1A1, CYP1B1, cytochrome P450, carcinogen metabolism, Mendelian randomization, machine learning, single-cell transcriptomics, oxidative stress, biomarkers, BEAS-2B

The review, authored by Shivam Kumar Vyas, Md Imran Ahmed, and Vaibhav A. Dixit at the National Institute of Pharmaceutical Education and Research in Guwahati, India, takes on the question embedded in its title: are cytochrome P450 inhibitors friends or foes? The authors’ answer is deliberately nuanced. Inhibition of specific CYP450 isoforms has emerged as a powerful therapeutic strategy in cancer and infectious disease, while unintended inhibition of drug-metabolizing isoforms remains a leading cause of clinically significant interactions. The challenge, they conclude, is not whether CYP450 inhibition is good or bad, but whether scientists can reliably predict which outcome a given inhibitor will produce.

The therapeutic case for inhibition is striking. CYP19A1, also known as aromatase, catalyzes the conversion of androgens to estrogens, and its blockade by drugs such as letrozole and anastrozole starves hormone-dependent breast cancer cells of the estrogen they need to proliferate. CYP17A1 plays a critical role in androgen biosynthesis, and its inhibition by abiraterone suppresses androgen production in prostate cancer. In the endocrine arena, inhibitors of CYP11B1 and CYP11B2, which drive corticosteroid and aldosterone synthesis, have been explored for Cushing’s syndrome and hypertension. Perhaps most famously, fungal CYP51 catalyzes a key demethylation step in ergosterol biosynthesis, and azole antifungals such as ketoconazole, itraconazole, and fluconazole exploit this dependency to collapse fungal cell membranes. Pathogen CYP450 enzymes have even been pursued as targets in tuberculosis and leishmaniasis.

The same chemistry, however, can turn hostile. Ketoconazole’s inhibition of CYP3A4 dramatically raises plasma levels of drugs like midazolam and simvastatin, producing excessive sedation or, in the case of statins, rhabdomyolysis risk. Fluconazole’s inhibition of CYP2C9 can elevate warfarin concentrations and precipitate bleeding complications. Prodrugs present a subtler trap: clopidogrel requires CYP2C19-mediated activation, and co-administration with inhibitors such as omeprazole blunts the formation of its active metabolite, diminishing antiplatelet efficacy and raising cardiovascular risk. Whether inhibition helps or harms, the authors argue, depends on isoform selectivity, the mechanism of inhibition, and — crucially — what metabolism actually does to the drug in question.

To bring order to a field crowded with overlapping terminology, the review proposes a unified classification. CYP450 inhibitors are sorted into two major mechanistic categories: reversible inhibitors and mechanism-based, or time-dependent, inhibitors. Reversible inhibitors, which are typically competitive and display Type I spectroscopic signatures, occupy the active site without undergoing oxidation. Mechanism-based inhibitors are subdivided into quasi-irreversible and irreversible types. Quasi-irreversible inhibitors coordinate the heme iron through a heterocyclic nitrogen or form metabolite-intermediate complexes, as seen with miconazole, paroxetine, and tadalafil. Irreversible inhibitors, by contrast, are bioactivated by the enzyme itself into reactive metabolites that covalently modify the heme or protein residues — a process the authors prefer to describe without the popular but imprecise label of suicide inhibition.

A key analytical contribution is the authors’ reassessment of kinetic behavior using a modified Michaelis-Menten parameter, β = Km/Ki, which captures the balance between a molecule’s role as a substrate and as an inhibitor of the same enzyme. When log β exceeds 1, non-productive inhibitory binding dominates and the enzyme effectively self-inhibits; when log β falls below 1, productive metabolism prevails. Applying this framework, phenacetin at CYP1A2 shows a log β of roughly 2.64, marking it as strongly inhibitor-like, while dextromethorphan at CYP2D6, with a log β near 0.19, behaves predominantly as a substrate. For mechanism-based inhibitors, the authors introduce a partitioning ratio, kinact/koff, derived from measurable equilibrium constants rather than hard-to-obtain rate constants, to quantify the relative contributions of reversible and irreversible binding. Miconazole’s ratio of 14 signals rapid irreversible binding, whereas tadalafil’s near-zero ratio indicates balanced contributions from both modes.

Spectroscopy provides an independent lens on these mechanisms. The resting ferric enzyme absorbs near 417–420 nanometers in its low-spin state. Type I ligands, which are substrate-like and displace the axial water ligand, shift the heme iron to a high-spin configuration and produce a characteristic difference spectrum with a peak near 385–390 nanometers and a trough near 418–425 nanometers. Type II ligands, typically nitrogen heterocycles such as imidazoles and triazoles, coordinate directly to the ferric heme iron, stabilize the low-spin state, and generate a red-shifted signature with a crest between 425 and 435 nanometers. Yet the authors caution that spectroscopic class and kinetic mechanism do not always align: ketoconazole, though classically Type II, exhibits mixed inhibition kinetics at CYP3A4 because the enzyme’s large, flexible active site can accommodate multiple ligands simultaneously.

On the regulatory side, the review walks through the FDA’s in vitro guidance for evaluating CYP450 substrates, inhibitors, and inducers, and proposes a streamlined workflow that integrates reversible and time-dependent inhibition testing into a single protocol. Reaction phenotyping with recombinant enzymes identifies which isoforms handle a compound; if a single enzyme accounts for more than 25 percent of metabolism, the compound is flagged as a sensitive substrate warranting further study. Inhibition screening at 10 to 30 micromolar, followed by IC50 determination and a preincubation-based IC50 shift assay, distinguishes reversible from time-dependent inhibitors. For reversible inhibitors, an R1 value below 1.02 clears the compound from further DDI concern; for time-dependent inhibitors, an R2 value below 1.25 serves the same gatekeeping function. The authors estimate their consolidated workflow could save four experimental steps compared with running the FDA’s separate protocols in parallel.

Perhaps the most practically useful contribution is a pair of decision criteria for predicting whether a given inhibitor will behave as a friend or a foe. Isoform selectivity, expressed as the difference in pIC50 between the therapeutic target and the nearest drug-metabolizing off-target, emerges as a strong predictor: a gap greater than one log unit — roughly tenfold selectivity — generally correlates with a favorable clinical profile. Letrozole, with a ΔpIC50 of 2.44, and ketoconazole, at 2.63, fit this pattern, while abiraterone, voriconazole, fluconazole, and itraconazole, with gaps at or below 0.75, are associated with significant clinical interactions. The second criterion concerns the direction of metabolism’s effect on drug action: when metabolism activates a prodrug or clears an active drug, inhibition tends to be harmful, as with ticagrelor plus ketoconazole or clopidogrel plus omeprazole; when metabolism inactivates the parent drug, inhibition can be harnessed as a pharmacokinetic booster, as ritonavir and cobicistat do for darunavir, atazanavir, lopinavir, and nirmatrelvir. Ritonavir itself illustrates the context-dependence, acting as a friend for some partners and a foe for others, including posaconazole and venetoclax.

The review closes with a survey of computational models developed over the past five years to predict CYP450 inhibition, ranging from random-forest classifiers like CYPlebrity to graph convolutional networks, deep neural networks such as CYPi-DNN, and multimodal architectures like the Multimodal Encoding Network, which combines molecular fingerprints, graphs, and protein sequences to achieve a reported 93.7 percent accuracy and 98.5 percent AUC. Yet the authors identify a conspicuous gap: no existing model directly predicts isoform selectivity, the very property their analysis shows is most predictive of clinical friend-or-foe behavior. Combining independently trained isoform-specific models, they note, is not a valid substitute because such combinations require their own validation. Until that gap is closed, the review’s central message stands: CYP450 inhibition is neither inherently friend nor foe, and the tools to tell the difference — kinetic, spectroscopic, and now computational — are finally converging on a coherent framework for making that call rationally.

Subject of Research: Mechanisms, classification, and therapeutic roles of cytochrome P450 enzyme inhibitors in drug metabolism and drug-drug interactions

Article Title: Cytochrome P450 inhibitors: friends or foes?

Article References: Vyas, S. K., Ahmed, M. I., & Dixit, V. A. (2026). Cytochrome P450 inhibitors: friends or foes?. Frontiers in Chemistry, 14, Article 1988344. https://doi.org/10.3389/fchem.2026.1988344

Image Credits: AI Generated

DOI: 10.3389/fchem.2026.1988344

Keywords: cytochrome P450, drug metabolism, drug-drug interactions, enzyme inhibition, pharmacokinetics, mechanism-based inhibition, CYP3A4, azole antifungals, machine learning, drug discovery, Type I and Type II ligands, pharmacokinetic boosting

The evidence base for this argument comes from resistance data collected between 2022 and 2023 across seven Rwandan districts. At sentinel sites including Bukora in Kirehe district, Gahini in Kayonza, Gashora in Bugesera, Kajevuba in Rulindo, Mubuga in Karongi, Nyagasambu in Rwamagana, and Rwanza in Musanze, Anopheles gambiae mosquitoes were collected during both rainy and dry seasons and tested with World Health Organization tube assays. The mosquitoes were exposed to two widely used pyrethroids, permethrin and alphacypermethrin, both alone and in combination with piperonyl butoxide, or PBO, a synergist that inhibits the cytochrome P450 enzymes mosquitoes use to detoxify insecticides.

The results were striking. Mortality from the pyrethroids alone was consistently below 80 percent at every site, and in some locations fell to between 40 and 50 percent—well below the WHO threshold of 98 to 100 percent mortality that defines a susceptible population. When mosquitoes were pre-exposed to PBO before the pyrethroid challenge, mortality rose significantly, indicating that a large share of the resistance is metabolic in origin, driven by overactive detoxification enzymes. Yet mortality still did not reach the 98 to 100 percent benchmark, which tells researchers something important: P450-mediated metabolic resistance explains only part of the problem, and additional resistance mechanisms must be operating in these populations.

Those additional mechanisms are diverse and, crucially, invisible to the standard bioassay. The policy brief lists target-site resistance—knockdown resistance mutations such as L1014F and L1014S in the voltage-gated sodium channel gene, the Ace-1 G119S mutation, and the N1575Y mutation—as likely contributors in some areas. Cuticular resistance, in which a thicker or chemically altered insect cuticle slows insecticide penetration, and behavioral resistance, in which mosquitoes shift their biting to earlier hours, move outdoors, or avoid sprayed surfaces, may also be present. Phenotypic testing with the WHO tube assay can confirm that resistance exists at a given site, but it cannot identify which mechanism is responsible, nor can it predict how resistance will spread over time. That diagnostic blind spot is the central technical argument of the brief.

This is where genomics enters the picture. The authors recommend integrating molecular and genomic tools into routine national entomological surveillance by 2027, including polygenic risk scores and targeted amplicon sequencing panels such as AgVampIR, which can screen mosquito populations for multiple resistance markers simultaneously. Such tools would allow the Rwanda Biomedical Centre’s Division of Malaria and Other Parasitic Diseases to detect emerging resistance earlier, track how resistance alleles move across the landscape, and build the evidence base for site-specific intervention decisions. The WHO itself supports routine entomological and insecticide-resistance surveillance using molecular and biochemical assays, emphasizing that accurate species identification and detection of resistance mechanisms require advanced molecular analysis.

The geographic heterogeneity documented in the 2022–2023 data carries a direct policy implication: interventions should be stratified by village or district rather than applied uniformly. Resistance levels vary significantly by district and ecological zone, and WHO guidelines recommend that surveillance be context-specific, reflecting those ecological differences. In practice, this means using village-level resistance profiles—including both the intensity of resistance and its molecular mechanisms—to select and rotate insecticides for each next malaria control campaign. The Rwanda Biomedical Centre would coordinate this at district, sector, and village levels, replacing the current one-insecticide-everywhere approach with a mosaic of locally matched interventions.

One immediate, evidence-backed recommendation is the prioritization of PBO-treated bed nets, or dual active-ingredient nets, in locations where metabolic resistance is confirmed. Because PBO restores susceptibility by blocking the detoxification enzymes, nets treated with the synergist perform better where P450-mediated resistance dominates. The brief calls for scaling up PBO net deployment in high-resistance villages and rotating non-pyrethroid insecticides where susceptibility allows, within two years, in line with WHO recommendations. The Ministry of Health, through the Rwanda Biomedical Centre, would handle procurement and distribution and establish an insecticide rotation schedule in collaboration with partners. The authors are careful to note, however, that PBO nets are not a permanent solution, since the residual mortality gap points to non-metabolic mechanisms that PBO cannot address.

Timing matters as much as tooling. Rwanda has two major malaria transmission seasons—the long rains from March to May and the short rains from September to December—which shape mosquito breeding and transmission intensity. The brief recommends expanding insecticide resistance surveillance to twice yearly at sentinel sites, collecting mosquitoes before and after the main rainy seasons to capture seasonal variation in mosquito abundance, insecticide exposure, and resistance levels. This cadence aligns with WHO guidance published in 2022 for routine, longitudinal resistance monitoring using consistent methods across different parts of the country, and it would give the national program an early-warning system for emerging resistance markers rather than a retrospective record of failure.

The economic framing of the brief is equally pointed. Citing the vision articulated by President Paul Kagame as the African Union champion for domestic health financing at the 34th African Union Summit, the authors argue that health investments should be judged by their impact on service coverage and health outcomes rather than by the amount spent. Indoor residual spraying, for example, may cost approximately two million dollars annually in a single district, and up to 24 million dollars when all target districts are included. If the insecticide deployed is ineffective against a resistant local mosquito population, that investment buys little protection. Strengthening resistance monitoring and targeting effective nets to the right locations, the authors argue, directly improves the efficiency of existing spending, reduces waste on ineffective interventions, and supports evidence-based resource allocation.

Finally, the brief stresses that insecticide resistance is not solely a public health problem. The same compounds used to kill mosquitoes are widely used in agriculture, and cross-sectoral pesticide use accelerates selection for resistance. The authors propose co-designing pesticide regulations with communities, coordinated by the health and agriculture sectors, and establishing by 2027 a joint task force spanning the Rwanda Agriculture and Animal Resources Development Board, the Rwanda Inspectorate, Competition and Consumer Protection Authority, the Ministry of Agriculture and Animal Resources, the Rwanda Environment Management Authority, the Rwanda FDA, the Rwanda Revenue Authority, the Rwanda Standards Board, and the University of Rwanda. The underlying message is that malaria elimination will not be achieved by traditional vector control tools alone: insecticide selection and rotation must be informed by regular genomic surveillance of mosquitoes, and pesticide stewardship demands collaboration among public health, agricultural, regulatory, and community stakeholders alike.

Subject of Research: Genomic surveillance of insecticide resistance in Anopheles gambiae mosquitoes to inform malaria vector control policy in Rwanda

Article Title: Genomic surveillance of mosquito resistance to guide cost-effective insecticide policy in Rwanda

Article References: Lambert, N., Dyer, N. A., Diane, U., Sherrard-Smith, E., & Leon, M. (2026). Genomic surveillance of mosquito resistance to guide cost-effective insecticide policy in Rwanda. Frontiers in Tropical Diseases, 7, Article 1887818. https://doi.org/10.3389/fitd.2026.1887818

Image Credits: AI Generated

DOI: 10.3389/fitd.2026.1887818

Keywords: Anopheles gambiae, insecticide resistance, genomic surveillance, malaria, Rwanda, pyrethroids, piperonyl butoxide, bed nets, indoor residual spraying, WHO tube assay, vector control, cytochrome P450

The research, conducted at the Chinese People’s Liberation Army General Hospital in Beijing and published in BMC Cancer, examined the records of pediatric patients with hematologic malignancies who developed PRES during chemotherapy between January 2011 and December 2024. Over that fourteen-year window, nineteen children met the criteria for the syndrome, and seventeen of them experienced seizures. The team, led by Lin-Yan Hu, Chen Feng, and senior authors Xiu-Yu Shi and Li-Ping Zou, focused on a specific clinical question: among children who needed ongoing antiseizure therapy after their acute episodes, which medication could be given without sabotaging the cancer treatment itself? Their answer, drawn from a small but carefully documented cohort, was topiramate, a drug with a pharmacological profile that largely sidesteps the metabolic machinery chemotherapy depends on.

To understand why this matters, it helps to consider the biology of the problem. Most chemotherapy agents are metabolized by the cytochrome P450 enzyme family in the liver, a system that also processes many older antiseizure medications. Drugs such as phenytoin, carbamazepine, and phenobarbital are potent inducers of these enzymes, meaning they can accelerate the breakdown of chemotherapeutic agents and reduce their effectiveness below therapeutic thresholds. In a child fighting acute lymphoblastic leukemia or Burkitt lymphoma, that interaction is not a theoretical concern; it can mean the difference between remission and relapse. Even newer agents like levetiracetam, which has become the default choice in many oncology settings because of its benign interaction profile, carry practical challenges, including behavioral side effects in children and dosing adjustments in kidney impairment, which is common in PRES itself.

Topiramate occupies an unusual niche in this landscape. Chemically a sulfamate-substituted monosaccharide, it dampens neuronal excitability through several mechanisms at once: it blocks voltage-gated sodium channels, enhances the inhibitory action of GABA at certain receptors, antagonizes the AMPA subtype of glutamate receptor, and modulates calcium channels. Crucially, it is largely eliminated unchanged by the kidneys rather than extensively metabolized by cytochrome P450 enzymes, which means it exerts minimal induction or inhibition on the pathways that clear chemotherapy drugs. This pharmacokinetic invisibility is precisely what makes it attractive for oncology patients, and it is the theoretical foundation on which the Beijing team built their clinical experience.

The study’s design was straightforward and descriptive. The researchers defined clear criteria for initiating maintenance antiseizure therapy: children who suffered more than two seizures within twenty-four hours, or more than three seizures within three days, were considered at sufficiently high risk of further episodes to warrant ongoing medication. Of the eleven children who met these thresholds, eight received topiramate at a median dose of three milligrams per kilogram per day, while three received levetiracetam at a median of thirty-five milligrams per kilogram per day. Six additional children who experienced only a single seizure did not receive maintenance therapy, reflecting a deliberate strategy of reserving long-term drugs for those with demonstrated recurrence risk.

The outcomes, though drawn from a small sample, tell a striking story. All eight children treated with topiramate completed their planned chemotherapy courses, and no serious adverse events attributable to antiseizure medications were documented anywhere in the cohort. One child in the topiramate group required the drug to be reintroduced after an initial discontinuation and went on to develop secondary epilepsy, a reminder that PRES can leave lasting neurological scars even when the acute imaging findings appear reversible. In the levetiracetam group, the picture was more sobering: one child died from cancer relapse after persistent neurological deficits made it impossible to tolerate further chemotherapy. Among the six children with single seizures who received no maintenance therapy, two likewise died of relapse after being unable to continue treatment.

These numbers must be interpreted with great care, and the authors are refreshingly candid about that. The study is retrospective, meaning the researchers looked backward at records rather than assigning treatments prospectively, and it was not randomized. The choice of topiramate versus levetiracetam was made by treating clinicians, likely influenced by each child’s specific circumstances, introducing selection bias that no statistical adjustment can fully remove. With only nineteen patients in total and just eleven on maintenance therapy, the study lacks the statistical power to demonstrate that topiramate improves survival or chemotherapy completion rates. The authors explicitly frame their findings as hypothesis-generating and descriptive, a scientific way of saying that the results are intriguing enough to justify further study but not definitive proof of benefit.

Nevertheless, the consistency of the signal is worth pausing on. In a population where the inability to tolerate chemotherapy is often the proximate cause of death, the observation that every topiramate-treated child completed their planned cancer treatment is clinically meaningful, even if causation cannot be established. The median dose used, three milligrams per kilogram per day, is modest by epilepsy standards, suggesting that seizure control in PRES may require less aggressive dosing than chronic epilepsy. The absence of serious adverse events is also notable, given that topiramate is known to cause cognitive slowing, kidney stones, and metabolic acidosis in some patients, side effects that could theoretically compound the toxicities of intensive cancer therapy.

The broader context makes this work part of a growing recognition that PRES in children with cancer deserves dedicated research rather than extrapolation from adult neurology. The syndrome arises when chemotherapy-induced hypertension, electrolyte disturbances, and direct endothelial toxicity overwhelm the autoregulatory capacity of the brain’s posterior circulation, leading to vasogenic edema visible on magnetic resonance imaging. Most children recover fully if the syndrome is recognized promptly and blood pressure is controlled, but the seizure burden during the acute phase can be substantial, and the question of how long to continue antiseizure drugs afterward remains unresolved worldwide. Guidelines are sparse, and practice varies widely between centers, which is precisely why even small descriptive studies like this one fill an important gap.

What happens next will determine whether topiramate’s pharmacokinetic advantages translate into real clinical benefit. The authors call for prospective validation, ideally a multicenter trial that randomizes children with PRES-related seizures to different maintenance regimens and tracks chemotherapy tolerance, seizure recurrence, and long-term neurological outcomes. Until such evidence exists, the Beijing experience offers pediatric oncologists and neurologists a reasoned option to consider when the interaction profile of standard antiseizure drugs threatens to undermine cancer therapy. It is a modest finding in a rare and complicated corner of medicine, but for the families of children facing both cancer and seizures simultaneously, even a modest finding can point toward a safer path through an already perilous journey.

Subject of Research: Use of topiramate as maintenance antiseizure therapy in children with hematologic malignancies who develop posterior reversible encephalopathy syndrome during chemotherapy

Article Title: Topiramate use for maintenance antiseizure therapy in children with hematologic malignancies experiencing posterior reversible encephalopathy syndrome during chemotherapy: a single-center retrospective study

Article References: Hu, L.-Y., Feng, C., Cai, R.-Y., Li, Y., Chai, L.-N., He, W., Chen, J., Wan, L., Yang, G., Shi, X.-Y., & Zou, L.-P. (2026). Topiramate use for maintenance antiseizure therapy in children with hematologic malignancies experiencing posterior reversible encephalopathy syndrome during chemotherapy: a single-center retrospective study. BMC Cancer. https://doi.org/10.1186/s12885-026-16999-9

Image Credits: AI Generated

DOI: 10.1186/s12885-026-16999-9

Keywords: posterior reversible encephalopathy syndrome, PRES, topiramate, levetiracetam, anti-seizure medications, pediatric oncology, hematologic malignancies, chemotherapy, seizures, drug interactions, cytochrome P450, retrospective study

The significance of the advance lies in what the weed represents. Herbicide resistance in weeds comes in two broad flavors. Target-site resistance arises from mutations in the proteins that herbicides attack, and these mutations are relatively easy to identify by sequencing known target genes. Non-target-site resistance, by contrast, is a murkier affair. It encompasses reduced herbicide uptake and movement through the plant, as well as enhanced metabolic detoxification, in which enzymes chemically dismantle the herbicide before it can do its work. Transcriptomic studies have catalogued hundreds of candidate genes associated with these processes, but cataloguing is not proof. Without a way to move candidate genes into the weed itself and observe the consequences, researchers have had to rely on indirect evidence from model species such as rice and Arabidopsis.

E. phyllopogon, a self-pollinating allotetraploid with 36 chromosomes, earned its reputation in California rice fields in the late 1990s, when populations were found resistant to herbicides spanning at least five different modes of action: acetolactate synthase inhibitors, acetyl-CoA carboxylase inhibitors, very-long-chain fatty acid elongase inhibitors, inhibitors of 1-deoxy-D-xylulose 5-phosphate synthase, and synthetic auxins. This broad-spectrum resistance is largely attributed to enhanced herbicide metabolism, a form of non-target-site resistance that has since been reported in an increasing number of weed species worldwide. Because metabolic resistance can confer cross-resistance to many herbicides at once, it poses a particularly serious threat to sustainable crop production, and E. phyllopogon has emerged as one of the best-studied systems for dissecting it.

Genomic and transcriptomic work has narrowed the search to a handful of suspects, most prominently three cytochrome P450 genes: CYP81A12, CYP81A21, and CYP709C69. These enzymes are strongly implicated in the detoxification of multiple herbicide classes in resistant populations, particularly a California-derived line known as 511. Intriguingly, the coordinated overexpression of these P450 genes appears to have been triggered by a single evolutionary event, possibly a regulatory mutation, which would explain how the weed acquired resistance to so many chemically unrelated compounds at once. But confirming that any given P450 actually detoxifies a given herbicide requires putting the gene into a living plant and watching what happens, something impossible without a transformation system.

The team, led by Takahiro Gondo of the University of Miyazaki together with Satoshi Iwakami of Tokyo University of Agriculture and Technology, built their protocol on methods previously developed for C4 warm-season forage grasses, including bahiagrass, ruzi grass, napier grass, and zoysiagrass. Their starting material was a herbicide-sensitive accession of E. phyllopogon collected from a paddy field in Nagano Prefecture, Japan. Seeds were sterilized and germinated, and after three days the basal tissues containing the shoot apical meristem were excised and placed on callus induction medium containing the auxin 2,4-dichlorophenoxyacetic acid, the cytokinin 6-benzylaminopurine, and copper sulfate. Within two weeks, primary calli emerged from the meristematic region, and by four weeks, embryogenic calli had formed at a frequency of roughly 20 to 30 percent.

A crucial early insight came from simply looking closely at the calli. The researchers classified them into three morphologically distinct types: Type 1 calli were compact and composed of clusters of somatic embryos; Type 2 calli were friable and highly proliferative, packed with immature embryogenic tissue; and Type 3 calli were wet and viscous. When the three types were tested for their ability to regenerate whole plants on medium containing naphthaleneacetic acid and kinetin, Types 1 and 2 produced approximately four shoots per callus and significantly outperformed Type 3. The differences carried through to gene transfer as well. After bombardment, Types 1 and 2 yielded roughly 70 transient GUS expression spots per callus, while Type 3 produced only about 2. The lesson was unambiguous: callus quality is a critical determinant of transformation efficiency, echoing findings from rice and maize, where actively proliferating embryogenic calli similarly boost success rates.

For the transformation itself, the team used a particle inflow gun, a device that accelerates DNA-coated gold particles into plant cells using a burst of helium. The plasmid pAHC25 carried two genes: a beta-glucuronidase reporter whose activity can be visualized by blue histochemical staining, and the bar gene, which confers resistance to bialaphos and its commercial formulation Basta. Both genes were driven by the maize polyubiquitin-1 promoter and its first intron, a combination known to drive strong expression in monocots. Before bombardment, the calli were pre-treated with mannitol and sorbitol to plasmolyze the cells, a step that reduces damage from the physical impact of the particles. Afterward, the calli were cultured on medium containing 5 milligrams per liter of bialaphos, and almost all bombarded calli gradually browned and died, while the rare transformed cells proliferated into visible resistant clumps within 90 days.

Timing proved to be the second key variable. Calli subcultured for 90 days delivered the highest transformation efficiency, 5.0 percent, and every putative transgenic callus from that age group successfully regenerated into a plant. In contrast, calli maintained for 300 and 630 days showed declining transformation efficiency, and most transgenic calli from these older cultures could not be regenerated at all. The researchers attribute this deterioration to changes in embryonic structure and loss of regenerative capacity over prolonged culture, along with an elevated risk of somaclonal variation, including albinism. Recent genome-wide studies have shown that tissue culture can substantially increase the accumulation of somatic mutations, raising genuine concerns about genomic integrity during clonal propagation. Minimizing culture duration, the team concludes, matters not only for regeneration but also for keeping the resulting plants genetically faithful to the source material.

The proof that the system works came from molecular and phenotypic verification. Southern blot analysis of five independent transgenic lines, using a GUS gene probe, confirmed stable integration of the introduced DNA, with copy numbers ranging from 5 to 12. Such multiple insertions are a well-known hallmark of particle bombardment, which tends to generate more complex integration patterns than Agrobacterium-mediated methods. Histochemical staining showed strong, stable GUS expression in the leaves of acclimatized transgenic plants. Most strikingly, when self-pollinated seeds from transgenic line #2 were grown and sprayed with Basta at the recommended field concentration, and then at ten times that concentration, the wild-type plants died completely while the transgenic progeny continued growing normally. The herbicide-resistant phenotype, in other words, was faithfully transmitted to the next generation.

Notably, the team first attempted Agrobacterium-mediated transformation, the method of choice in many grasses, but in preliminary experiments with E. phyllopogon they detected neither infection nor transient GUS expression. Agrobacterium efficiency is notoriously species- and genotype-dependent, and this weed proved recalcitrant, so direct DNA delivery via bombardment became the path forward. The result is the first report of transgenic E. phyllopogon plants, a platform the researchers say will enable functional analysis of candidate genes involved in herbicide metabolic detoxification, including the CYP81A and CYP709C P450s, and will help unravel how a single regulatory event can lock broad-spectrum resistance into a weed’s genome. For a species that has resisted both herbicides and genetic manipulation, the resistance to the latter has finally been broken.

Subject of Research: Development of the first particle bombardment-mediated genetic transformation system for the herbicide-resistant weed Echinochloa phyllopogon

Article Title: Development of a genetic transformation system in Echinochloa phyllopogon by particle bombardment

Article References: Gondo, T., Sugahara, M., Ng, H. M., Yasuda, K., & Iwakami, S. (2026). Development of a genetic transformation system in Echinochloa phyllopogon by particle bombardment. Plant Cell Reports, 45(11), Article 331. https://doi.org/10.1007/s00299-026-03978-4

Image Credits: AI Generated

DOI: 10.1007/s00299-026-03978-4

Keywords: Echinochloa phyllopogon, genetic transformation, particle bombardment, herbicide resistance, non-target-site resistance, cytochrome P450, embryogenic callus, bialaphos, GUS reporter, plant regeneration, rice paddy weed, metabolic detoxification