MicroRNAs, or miRNAs, are short non-coding RNA molecules, typically only about 22 nucleotides long, that do not encode proteins. Instead, they act as fine-tuners of gene expression. First discovered in the roundworm Caenorhabditis elegans and since found across most eukaryotes, including humans, a single miRNA can regulate the activity of multiple genes simultaneously. This one-to-many regulatory capacity makes them especially interesting for complex, multi-gene disorders such as cancer. Because miRNAs show distinct expression patterns in healthy individuals, cancer patients, and even across different disease stages, scientists have increasingly viewed them as candidates for diagnostic and prognostic indicators, and in some cases as drug targets in their own right.

In the new study, Anushka Pravin Chawhan and Norine Dsouza of St. Xavier’s College in Mumbai set out to close a gap in the LUSC literature. While many previous studies had catalogued differentially expressed protein-coding genes in this cancer, miRNAs had not been broadly exploited as targets. The researchers mined open-access miRNA sequencing data from The Cancer Genome Atlas, or TCGA, applying filters for transcriptome profiling, miRNA-Seq, and miRNA expression quantification. They downloaded and processed the datasets using the R package TCGAAssembler2, then compared miRNA expression between primary solid tumour samples and adjacent normal tissue to find molecules behaving abnormally in cancer.

The differential expression analysis, carried out in RStudio with the DESeq2 package, produced a strikingly balanced picture. Using an adjusted p-value threshold of less than 0.05, the team identified 81 differentially expressed miRNAs in LUSC tumours. Of these, 41 were down-regulated, consistent with a tumour-suppressing role, while 40 were up-regulated, suggesting oncogenic behaviour. The classification rested on the log2 fold change in expression, with values below -0.1 marking potential tumour suppressors and values above 0.1 marking potential oncogenes. A volcano plot generated with ggplot2 visualised the split, and the full list of significant molecules was deposited in supplementary files accompanying the paper.

Identifying molecules that behave differently in tumours is only half the story, however. The more clinically urgent question is whether those differences actually matter for patients. To answer it, the authors performed Cox proportional hazards regression on 1,519 miRNAs, combining tumour miRNA expression data with clinical outcomes from TCGA. Patients were stratified into high-expression and low-expression groups for each miRNA, and a log-rank p-value below 0.05 flagged molecules with a significant impact on overall survival. The analysis revealed another 81 survival-related miRNAs: for 46 of them, low expression predicted better survival, while for 35, high expression was associated with improved outcomes.

The pivotal moment came when the two lists were compared. Seven miRNAs appeared in both: they were differentially expressed in tumours and simultaneously had a significant effect on how long patients lived. These seven candidates were miR-129-2, miR-181a-1, miR-501, miR-519a-1, miR-545, miR-6509, and miR-6761. The survival directions differed among them. High expression of miR-129-2, miR-501, and miR-6509 correlated with better overall survival, whereas low expression of miR-181a-1, miR-519a-1, miR-545, and miR-6761 was linked to the same favourable outcome. Each of the 74 remaining miRNAs in each list was unique to one analysis, underscoring how stringent the double filter really is.

To probe what these seven molecules actually do, the team turned to the MISIM v2 web tool, which infers miRNA functional similarity based on known miRNA-disease associations. Building an interaction network with only positive correlations between 0.5 and 1, they found that five of the seven candidates, namely miR-129-2, miR-181a-1, miR-501, miR-519a-1, and miR-545, formed an interconnected cluster. Functional enrichment analysis tied the group to critical biological processes including cell death, blood-tumour barrier permeability, chondrocyte development, chondrogenic differentiation, osteoclast genesis, and vascular inflammation, pointing to deep involvement in cell differentiation machinery.

The disease association analysis broadened the picture even further. The candidate miRNAs were predicted to be involved in a remarkable range of malignancies, including breast, gastric, hepatocellular, colon, prostate, thyroid, renal cell, urothelial, and cervical cancers, as well as lung adenocarcinoma, lung neoplasms, brain neoplasms, and chondrosarcoma. Intriguingly, the molecules were also linked to neurological conditions such as neuroinflammation and Parkinson’s disease, a connection the authors note follows naturally from their involvement in brain tumours. This cross-disease web suggests the seven miRNAs sit at regulatory hubs whose perturbation reverberates through many tissue types, which is precisely the kind of biology that makes a molecule valuable as a biomarker.

Much of the study’s credibility comes from how well its computational findings align with prior experimental work. High expression of miR-6509 has been shown to reduce proliferation and migration and increase apoptosis in hepatocellular and gastric cancer cell lines, and ovarian cancer patients with high miR-6509 enjoy better five-year survival. miR-501 has repeatedly emerged as a tumour suppressor in renal, lung, and prostate cancers, restricting tumour size and metastasis. miR-129-2 is down-regulated in several cancers by targeting genes such as SOX4, BZW1, and Wip1, and its over-expression drives tumour reduction, consistent with the better survival seen in high-expression LUSC patients here. miR-181a-1, by contrast, behaves as a tumour promoter in colorectal cancer, multiple myeloma, and non-small cell lung cancer, matching the study’s finding that low expression predicts better outcomes.

Not every candidate fits neatly into the existing literature. miR-545 presents a genuine contradiction: one study found high expression lowers cell viability in lung adenocarcinoma and LUSC cell lines, yet the same work reported that the molecule promotes proliferation in HFL1 lung fibroblasts, cells that themselves can promote non-small cell lung cancer. The authors call for further clarity on this molecule, even as they note it has been validated as a biomarker in oncological studies. miR-519a shows similarly mixed signals, with high expression linked to poorer survival in liver and breast cancer, and one non-small cell lung cancer study, limited to adenocarcinoma cell lines, contradicting the present findings. Most tantalising of all, miR-6509 and miR-6761 had never before been implicated in LUSC, and the authors propose their involvement for the first time.

The authors are careful about what their results can and cannot claim. Everything rests on computational analysis of public datasets, and they emphasise that experimental validation is essential before any clinical translation. Knockdown and overexpression studies, in vitro experiments, animal models, and eventually clinical trials would all be needed to confirm efficacy, safety, and practical applicability, and to establish protocols for real-world patient care. Still, the logic of the approach is compelling: by demanding that a molecule be both dysregulated in tumours and statistically tied to patient survival, the study filters hundreds of candidates down to seven with genuine clinical potential. If laboratory work bears out the computational signals, these microRNAs could eventually feed into diagnostic panels for earlier detection, prognostic tools for personalised risk assessment, and even RNA-based therapeutics, offering a new molecular foothold against one of medicine’s most stubborn cancers.

Subject of Research: Computational identification of survival-associated microRNAs in lung squamous cell carcinoma

Article Title: Identification of key microRNAs in lung squamous cell carcinoma: a computational study

Article References: Chawhan, A. P., & Dsouza, N. (2025). Identification of key microRNAs in lung squamous cell carcinoma: a computational study. Discover Biotechnology, 2(1), Article 4. https://doi.org/10.1007/s44340-025-00011-4

Image Credits: AI Generated

DOI: 10.1007/s44340-025-00011-4

Keywords: microRNA, lung squamous cell carcinoma, TCGA, biomarkers, differential expression, survival analysis, Cox regression, non-small cell lung cancer, bioinformatics, oncogenes, tumor suppressors, precision medicine

The physical chemistry underlying condensates is elegant in its simplicity. Multivalent molecules, particularly proteins containing intrinsically disordered regions and repeated interaction motifs, can demix from the surrounding nucleoplasm or cytoplasm when their local concentrations and interaction energies cross a threshold, much like oil droplets forming in water. Early landmark studies demonstrated that germline P granules behave as liquids that localize by controlled dissolution and condensation, and subsequent work established the nucleolus as a multiphase liquid condensate and showed that super-enhancers assemble transcriptional coactivators into phase-separated compartments that link genome organization to gene control. This nucleation landscape is exquisitely sensitive to molecular composition, salt concentration, temperature, pH and post-translational modifications, meaning the cell can tune condensate formation and dissolution with impressive speed and reversibility.

Cancer exploits this tunability. The review synthesizes extensive evidence that aberrant LLPS contributes to tumorigenesis across a striking range of mechanisms. Oncogenic fusion proteins, which arise from chromosomal translocations and are common drivers of sarcomas, leukemias and brain tumors, frequently acquire prion-like low-complexity domains that promote ectopic phase separation, retargeting chromatin regulatory complexes and rewiring transcriptional programs. The BRD4-NUT fusion oncoprotein forms aberrant condensates that hijack the transcriptional machinery in NUT carcinoma, while EML4-ALK condensates in lung cancer and phase-separating NTRK fusions concentrate kinase signaling into cytoplasmic granules that fire oncogenic pathways constitutively. Disease-associated mutations in the phosphatase SHP2 promote phase separation that underlies MAPK hyperactivation, and point mutations in the chromatin reader ENL create condensate-promoting hotspots that drive leukemogenesis in vivo.

Tumor suppressors, too, are entangled with condensate biology, sometimes in paradoxical ways. The tumor suppressor SPOP normally organizes active phase-separated compartments in the nucleus, and cancer mutations disrupt these assemblies, impairing the destruction of oncogenic substrates. The histone demethylase UTX exerts its anti-tumor activity at least in part through condensation, and perturbing TET2 condensation alters genome-wide DNA methylation patterns in leukemia cells. On the oncogenic side, transcriptional co-activators YAP and TAZ, key effectors of the Hippo pathway, use phase separation to compartmentalize transcription machinery, reorganize genome topology and sustain long-term target gene expression, and interferon-gamma-driven YAP condensation has even been implicated in tumor resistance to anti-PD-1 immunotherapy. Stress granules themselves can act as oncogenic platforms, with RIOK1 phase separation restricting PTEN translation in hepatocellular carcinoma and PABPC1 condensates controlling selective translation in chronic myeloid leukemia blast crisis.

The reach of dysregulated phase separation extends into the machinery that governs innate immunity, an interface of growing therapeutic importance. The DNA sensor cGAS undergoes liquid phase condensation upon binding cytosolic DNA, activating type I interferon signaling that can either suppress tumors or, when subverted, promote tumorigenesis. Mutant NF2 induces phase separation that imprisons the cGAS-STING machinery, abrogating anti-tumor immunity, while STING itself acts as a phase-separator that can suppress innate immune signaling. Hypoxia, a defining feature of tumor microenvironments, reshapes the condensate landscape: low oxygen triggers ZHX2 phase separation that alters chromatin looping to drive metastasis, promotes FUS-circRNA stress granules that fuel autophagy in triple-negative breast cancer, and lactate-sensitive enzymes such as AARS1 modify cGAS and p53 through lactylation, tilting condensate dynamics toward tumor progression. Even metabolic states matter, as glycogen accumulation and phase separation have been shown to drive liver tumor initiation.

What elevates this review from a catalogue of mechanisms to a therapeutic manifesto is its detailed treatment of engineering strategies designed to modulate condensates directly. Among the most inventive are D-amino-acid-based interfering peptides, which exploit the protease resistance of mirror-image peptides to infiltrate and disrupt pathological condensates. Short designer peptides have already been shown to disassemble tau fibrils in models of neurodegeneration, and the same logic is being applied to oncogenic droplets. Pharmacological inhibition of SRC-1 phase separation suppresses YAP-driven transcription, targeting androgen receptor phase separation can overcome resistance to antiandrogen therapies in prostate cancer, and dissolution of oncofusion transcription factor condensates has emerged as a viable strategy for fusion-driven malignancies. Recently, targeting FOXM1 condensates reduced breast tumor growth and metastasis in preclinical models, and disruption of the KAT8-IRF1 condensate diminished PD-L1 expression, thereby promoting anti-tumor immunity.

Post-translational modifications offer a second, highly programmable axis of control. Phosphorylation, acetylation, methylation, ubiquitination, SUMOylation, PARylation, O-GlcNAcylation and lactylation each tune the multivalent interactions that drive condensate assembly or dissolution. RNA polymerase II C-terminal domain hyperphosphorylation is itself governed by a phase-separation mechanism, and phosphorylation of HDAC6 drives aberrant chromatin architecture in triple-negative breast cancer. PARP1-mediated poly(ADP-ribosylation) can disrupt condensates to halt global transcription after DNA damage, while sirtuin-sensitive acetylation of TDP-43 drives its pathological phase separation. The review argues that post-translational modification-mimetic approaches, in which engineered molecules co-opt these chemical switches, could allow clinicians to flip condensate states in tumor cells with a specificity that traditional enzyme inhibitors have struggled to achieve against disordered proteins.

Perhaps the most forward-looking section concerns synthetic condensates as drug delivery platforms. Rather than merely dissolving harmful droplets, bioengineers are learning to build benign ones. In situ formation of biomolecular condensates can create intracellular drug reservoirs that augment chemotherapy, while coacervate vesicles assembled through LLPS improve the delivery of biopharmaceuticals, and phase-separating peptides enable direct cytosolic delivery of macromolecular therapeutics with redox-triggered release. Programmable synthetic condensates have been used to enhance translation from target mRNAs and to control cellular behavior, and intrinsically disordered region-induced condensation has improved the cytotoxicity of CAR-T cells against low-antigen cancers. Design of intrinsically disordered region-binding proteins and micropeptide killswitches that probe condensate microenvironments point toward an era in which the chemical milieux inside droplets—differences in polarity, pH and redox state—can be navigated to sharpen drug targeting and overcome resistance.

The authors are candid about the obstacles that separate this vision from routine clinical practice. Condensates lack defined binding pockets, complicating conventional structure-guided drug design, and the same material properties that make them dynamic also make them hard to model computationally. Advances in machine learning predictors such as catGRANULE 2.0, STARLING and PSPire, alongside phase-separation-directed screening and improved microscopy methods, are beginning to close the gap by identifying which condensates exist in a given tumor and which modulators shift their dynamics. The reviewers also emphasize translational design considerations, including how therapeutics partition into nuclear condensates, a factor shown to influence the efficacy of cancer drugs. By integrating biophysical insight with engineering platforms, Wang, Zhang and Zhou outline a conceptual framework in which oncogenic drivers long dismissed as undruggable—disordered transcription factors, fusion oncoproteins, scaffold proteins—can finally be engaged through the physics of their assemblies. If that framework holds up in the clinic, the liquid droplets that cancer co-opted for its own ends may become the very vulnerabilities that defeat it.

Subject of Research: The role of liquid–liquid phase separation and biomolecular condensates in cancer pathogenesis and therapy

Article Title: Biomolecular condensates in cancer therapy

Article References: Wang, B., Zhang, L., & Zhou, F. (2026). Biomolecular condensates in cancer therapy. Nature Reviews Bioengineering. https://doi.org/10.1038/s44222-026-00493-9

Image Credits: AI Generated

DOI: 10.1038/s44222-026-00493-9

Keywords: biomolecular condensates, liquid-liquid phase separation, cancer therapy, LLPS, oncogenic fusion proteins, interfering peptides, synthetic condensates, drug delivery, post-translational modifications, tumor suppressors, Hippo-YAP signaling, bioengineering