The answer, at broad taxonomic levels, appears to be yes. The research team collected matched samples from patients undergoing treatment for bladder cancer under strictly sterile conditions, obtaining 79 bladder tissue specimens and 79 catheterized urine specimens, along with 17 rectal-swab stool samples from a subset of the same patients. Rather than the older 16S approach, the researchers deployed shotgun metagenomic sequencing, which reads all of the genetic material in a sample and can identify microbes at species-level resolution while also revealing the functional genes they carry. This matters because knowing which organisms are present is only half the story; the other half is understanding what metabolic pathways those organisms might be executing, information that species-level resolution makes possible to infer.

To analyze the data rigorously, the team divided patients into two groups based on specimen availability. The M-TUS group comprised 17 patients from whom tissue, urine, and stool were all available, while the larger M-TU group of 62 patients contributed matched tissue and urine only. The researchers then compared the three sample types using a battery of standard microbiome metrics. Alpha diversity measures, which capture the richness and evenness of species within each sample, showed that tissue and urine microbiomes were broadly similar to one another in both respects. When the team turned to beta diversity, assessed through Bray-Curtis dissimilarity, a metric that quantifies how different two microbial communities are in composition, the pattern was clear: tissue and urine samples were more similar to each other than either was to stool, which emerged as biologically distinct from the other two compartments.

That distinction is important for interpreting the growing literature on microbiomes and cancer. Stool samples are far easier to collect than tissue biopsies, and many studies have used gut microbiota as a convenient window into the body’s microbial state. But this study demonstrates that the gut community does not stand in for the bladder community. Species-sharing analysis reinforced the point, showing that tissue and urine samples shared more of their microbial species with each other than either shared with stool. The researchers extended the comparison to function as well, using metagenomically inferred metabolic pathways to ask whether the predicted functional potential of the tissue and urine communities converged. It did: predicted functional pathway concordance further supported the tissue-urine relationship and underscored the biological separateness of the gut compartment.

The team also probed whether microbial profiles differed by patient characteristics, specifically sex and tumor invasion status, which distinguishes non-muscle-invasive bladder cancer from the more aggressive muscle-invasive form. Class- and genus-level profiles were broadly similar between the sexes, and the differences between invasion groups were minimal. One apparent association between urine microbiota and sex did surface in the PERMANOVA statistical test, a permutational multivariate method widely used in microbiome research. However, the accompanying PERMDISP analysis, which tests whether groups differ simply in the spread of their samples rather than in their composition, revealed significant dispersion heterogeneity. The authors therefore interpreted the sex-associated result cautiously, a methodologically sound decision that reflects a common pitfall in microbiome statistics, where differences in within-group variability can masquerade as genuine compositional differences between groups.

Perhaps the most ambitious part of the study was an attempt to build a predictive model: could the team take a urine sample and computationally reconstruct the microbial profile of the corresponding bladder tissue? The researchers used ridge regression, a regularized linear modeling technique well suited to high-dimensional biological data, combined with principal component analysis to reduce dimensionality before fitting. The model was trained on the larger M-TU cohort and then tested on the independent M-TUS cohort, providing a genuine held-out validation. The results were strikingly level-dependent. At the phylum level, the model achieved an R-squared of 0.907 on the test set, meaning it explained over ninety percent of the variance in tissue microbial relative abundances from urine data alone. At the class level, performance remained strong with an R-squared of 0.834.

But the predictive power eroded as the taxonomic resolution sharpened. At the family and genus levels, the ridge regression model failed to outperform a null model, indicating that urine cannot yet reliably predict which specific families or genera populate the bladder tissue. This gradient makes biological sense: broad taxonomic categories capture the dominant structural features of a community, which are more likely to be shared between the urine and the tissue it contacts, whereas finer-grained composition may be shaped by local microenvironments on the bladder wall, differential shedding rates, and subtle sampling effects that a urine specimen cannot fully capture. The authors are candid about this limitation, noting that fine-resolution prediction will require validation in larger, multi-institutional, multi-ethnic cohorts that also include healthy controls, which the current cancer-only design lacked.

The practical implications are nonetheless considerable. Bladder cancer is one of the most common malignancies worldwide, and patients typically face lifelong surveillance because of the disease’s high recurrence rate. Current monitoring relies on cystoscopy, an invasive procedure, and urine-based biomarkers such as nuclear matrix protein 22, none of which capture microbial information. If catheterized urine can serve as a faithful proxy for tissue microbiota at broad taxonomic levels, as this study supports, it opens the door to less invasive microbiome profiling and, crucially, to longitudinal monitoring. A clinician could track a patient’s bladder microbial community over time through repeated urine collections, watching for shifts that might accompany recurrence, progression, or response to intravesical therapies, without ever touching the bladder wall with a scope.

The study also carries a methodological message for the field. By combining sterile catheterized collection with shotgun metagenomics, the researchers sidestepped the two biggest confounders of earlier urinary microbiome work: contamination from adjacent anatomical sites and the coarse resolution of 16S amplicon sequencing. The demonstration that stool is compositionally and functionally distinct from both urine and bladder tissue serves as a caution against extrapolating gut microbiome findings to urological contexts. At the same time, the honest reporting of where the predictive model broke down, at family and genus levels, illustrates the kind of transparency that builds trust in microbiome science. The work was supported by the National Research Foundation of Korea and other Korean funding bodies, with samples prepared through the National Biobank of Korea at Chungbuk National University Hospital, and it stands as a careful, technically rigorous step toward making the urinary microbiome a usable clinical instrument rather than an intriguing curiosity.

Subject of Research: Comparative microbiota analysis of urine, bladder tissue, and stool in bladder cancer using shotgun metagenomics

Article Title: Comparative analysis of urinary, bladder tissue, and stool microbiota in bladder cancer using shotgun metagenomics

Article References: Comparative analysis of urinary, bladder tissue, and stool microbiota in bladder cancer using shotgun metagenomics. (n.d.). https://doi.org/10.1186/s12916-026-05164-5

Image Credits: AI Generated

DOI: 10.1186/s12916-026-05164-5

Keywords: bladder cancer, microbiome, microbiota, shotgun metagenomics, urine, bladder tissue, stool, ridge regression, Bray-Curtis dissimilarity, biomarker, PERMANOVA, BMC Medicine

Urban soils are workhorses of the modern city. They buffer floods, counter the urban heat island effect, capture and degrade pollutants, support infrastructure, enable urban agriculture, and provide the green spaces that benefit both physical and mental health. Many of these ecosystem services depend on soil microbial communities, which drive nutrient cycling, decompose organic matter, and sustain plant health. There is even growing evidence that exposure to natural microbial biodiversity can benefit the human immune system. Yet while the microbiomes of forests and farmland have been studied extensively, urban soil microbes have remained strikingly underexplored, largely because urban soils are notoriously difficult to characterize. Construction, sealing, soil compaction, debris burial, and pollution from traffic and industry all scramble the physical, chemical, and biological properties of city ground in ways that vary block by block.

To untangle this complexity, the researchers designed an unusually rigorous sampling scheme. They selected 20 sites across Antwerp, each anchored by solitary London plane trees (Platanus × acerifolia) growing in unsealed soil strips next to paved streets. Using a single tree species kept the vegetation effect consistent across sites. At each site, soils were sampled under two trees spaced 3 to 20 meters apart; under each tree, samples were taken on the north and south sides, 1 to 3 meters apart; and at each orientation, the upper mineral soil was collected at two depths, from 0 to 5 centimeters and from 10 to 15 centimeters. In total, 160 soil samples were gathered between mid-February and the end of March 2023, spanning spatial distances from 2 meters to 7 kilometers.

The molecular work was equally ambitious. DNA was extracted from every sample and the bacterial 16S rRNA gene was amplified and sequenced on an Illumina MiSeq platform. After stringent quality filtering with the DADA2 pipeline, which discards low-quality reads and removes chimeric sequences, the final dataset contained roughly 1.3 million reads across 122 samples, resolving 11,093 amplicon sequence variants representing 1,835 bacterial genera. The dominant phyla were familiar players from natural soils: Proteobacteria accounted for about 24 percent of reads, followed by Acidobacteriota at 20 percent, Actinobacteriota at 19 percent, Firmicutes at 7 percent, and Verrucomicrobiota at 6 percent. Genera such as Renibacterium, Bradyrhizobium, Nitrosocosmicus, and Hyphomicrobium were among the most abundant, hinting at active carbon, nitrogen, and pollutant cycling even in this heavily engineered environment.

Alongside the sequencing, the team measured a battery of soil and environmental variables. Gravimetric moisture content averaged 15 percent, notably drier than the roughly 21 percent reported for nearby agricultural soils, consistent with previous observations that urban soils tend to be parched. Soil pH, measured in potassium chloride, had a median of 6.4, and a striking 93 percent of samples were more alkaline than the median pH of 4.6 typical for comparable forest soils in the same region, a signature of calcareous leaching from concrete. The researchers also measured soil density and saturation isothermal remanent magnetization, a magnetic proxy for traffic-derived heavy metal pollution, and pulled modelled concentrations of particulate matter and nitrogen oxides from high-resolution air quality maps. Land use classes around each site were extracted from the Copernicus Urban Atlas.

When the team ran permutational multivariate analyses of variance on the community data, a clear hierarchy emerged. The single strongest driver of bacterial community composition was the sampling site itself, explaining 40 percent of the variation, far outweighing any soil characteristic. The identity of the individual tree added another 13 percent, and soil depth and orientation contributed further significant effects. Soil properties did matter: pH, moisture, density, depth, air pollution, and land use class were all significantly associated with community composition, with pH emerging as the most influential soil variable. But its explanatory power was an order of magnitude smaller than that of spatial location. The researchers speculate that this overwhelming site effect reflects dispersal limitation combined with the historic movement of soil during construction, meaning the microbial inhabitants of whatever soil was originally dumped or disturbed decades ago still dominate the community today.

Two findings stand out for their practical implications. First, bacterial diversity dropped significantly as soil moisture declined, a relationship that became especially steep in the drier soils that dominate the city. Since urban soils are already prone to drying, this suggests that simple interventions, from removing sealing and installing permeable pavements to adding organic ground covers and improving water retention, could help preserve the microbial diversity that underpins urban soil function. Second, the age of the vegetation left a clear fingerprint on community dynamics. Soils under trees with smaller trunks, and therefore presumably younger trees and less developed root systems, varied enormously in bacterial composition, while soils under larger, older trees converged toward significantly more similar communities. The influence of pH and heavy metal pollution on community composition also weakened as trunk circumference increased, with the explained variation for pH falling from around 0.19 to 0.09 in surface soils. The interpretation is tantalizing: as trees and their root systems mature, the associated bacterial communities stabilize and may become more resilient to chemical stress.

The study also confirmed patterns familiar from natural ecosystems. Bacterial diversity peaked at a near-neutral pH between 6 and 7, echoing classic findings from continental-scale surveys of non-urban soils. But the urban context adds a twist: because concrete alkalizes city soils, much of the urban pH range sits above this optimum, and elevated pH has been linked to reduced microbial function. A global study of 56 cities previously found that urban green space microbiomes lean toward faster nutrient cycling, more greenhouse gas emissions, and greater pathogenic potential compared with nearby natural soils, and the Antwerp results add a crucial nuance: within a single city, the legacy of soil disturbance may matter as much as any global trend.

The authors are careful about the limits of their exploratory, single-time-point design. They sampled only bacteria, leaving fungi, archaea, protists, and other eukaryotes unexamined, and they could not directly observe successional dynamics without sampling over multiple years. They also note that without a proper definition of what constitutes a healthy urban soil microbiome, management recommendations remain partly hypothetical. Still, the message for urban planning is hard to ignore. Soil moisture, pH, vegetation age, and the history of ground disturbance all leave measurable signatures on the bacterial communities that deliver urban ecosystem services. As cities expand and redevelop, the researchers argue, microbial soil diversity deserves a seat at the planning table, alongside permeable surfaces, moisture-retaining designs, and conservation of older, undisturbed soils. The ground beneath our streets, it turns out, is not dead matter but a living archive of the city’s construction history, and learning to read and manage it may be one of the most underappreciated opportunities in modern urban sustainability.

Subject of Research: Urban soil bacterial diversity and its relationship with soil physicochemical properties in Antwerp

Article Title: Heterogeneity of the urban soil microbiome and associations with physicochemical soil characteristics

Article References: Smets, W., Bosiers, T., De Groof, N., Diels, E., Dubois, J., Gilis, F., Santullo Latorre, A., & Wuyts, K. (2026). Heterogeneity of the urban soil microbiome and associations with physicochemical soil characteristics. Web Ecology, 26(1), 35-45. https://doi.org/10.5194/we-26-35-2026

Image Credits: AI Generated

DOI: 10.5194/we-26-35-2026

Keywords: urban soil microbiome, soil bacteria, Antwerp, 16S rRNA sequencing, soil pH, soil moisture, London plane trees, urban planning, ecosystem services, soil pollution, PERMANOVA, microbial diversity

The research team, led by scientists at the Department of Geology of the University of Dhaka, collected 34 sediment samples along a land-sea transect spanning three environmentally contrasting areas: Cox’s Bazar on the eastern coast, Kuakata in the central deltaic zone, and the Sundarbans mangrove forest in the west. Sampling took place during winter, within the intertidal zone at elevations of roughly 0.5 to 4 meters above mean sea level, with cores extracted to a depth of 30 centimeters using a hand auger at low tide. Each sample was dried at 60 degrees Celsius for 24 hours, and 100-gram aliquots were sieved through meshes ranging from 63 to 1000 micrometers at one-phi intervals, following the Wentworth grain-size scale. Mean grain size and sorting were calculated using the Folk and Ward method.

The three study areas could hardly be more different. Cox’s Bazar, home to one of the longest natural beaches on Earth at 145 kilometers, is dominated by fine and medium sand, with fine sand reaching up to 75.1 percent of a sample and averaging 48.2 percent, while mud fractions are scarce. Kuakata displays a mixed picture: sediments at the Tin Nodir Mohona river mouth average 62.4 percent coarse sand, whereas the nearby Jhau Bon and Lebu Bon sites are rich in fine and very fine sand. The Sundarbans, the world’s largest mangrove forest covering roughly 6017 square kilometers, is a low-energy setting where fine and very fine sand dominate, averaging 59.5 and 27.2 percent respectively. These textural contrasts reflect fundamentally different hydrodynamic regimes, from wave-dominated open coast to tide-swept estuary to sheltered, freshwater-influenced mangrove plain.

When the researchers examined the foraminifera preserved in these sediments, the diversity patterns mapped almost perfectly onto the sediment map. Cox’s Bazar yielded 36 species across 26 genera, the highest diversity of the three areas, with Fisher’s alpha index ranging from 9.95 to 17.06 and Shannon values between 2.71 and 3.23. Kuakata produced 32 species from 20 genera with intermediate indices, while the Sundarbans assemblage contained only 19 species from 14 genera, the lowest diversity, with Fisher’s alpha falling as low as 4.89. The single most diverse site was Inani beach in Cox’s Bazar, and the least diverse was Kotka in the Sundarbans, a contrast that the authors link directly to substrate texture and the ecological conditions it implies.

The composition of the assemblages was equally revealing. The genera Ammonia and Elphidium dominated throughout the entire coastline, demonstrating their remarkable ecological tolerance, but each region carried a distinctive signature of companion taxa. Cox’s Bazar was characterized by Rotalinoides, Ammonia, Elphidium, and Asterorotalia, with mean relative abundances of 17.9, 17.4, 12.8, and 9.3 percent respectively. Kuakata was overwhelmingly dominated by Elphidium and Cribroelphidium at 33.8 and 20.7 percent, alongside several Ammonia species. In the Sundarbans, Ammonia reached individual-sample abundances of up to 55.6 percent, accompanied by Rosalina, Haplophragmoides, Entzia, Haynesina, and Trochammina, a suite of stress-tolerant and salt-marsh agglutinated forms typical of fine-grained, organically enriched, freshwater-influenced environments.

To test whether these patterns were statistically robust rather than coincidental, the team deployed a battery of complementary analyses. Spearman’s rank correlation across the combined 34-sample dataset showed that 18 of the 30 analyzed genera had at least one statistically significant association with a sediment fraction. Medium sand emerged as the fraction with the strongest biological signal: twelve genera, including Amphistegina, Spiroloculina, Cibicides, Rotalinoides, Neorotalia, and Operculina, showed significant positive correlations with it, while Ammonia, Elphidium, Rosalina, and Haplophragmoides were significantly and negatively correlated, reflecting their preference for finer substrates. The genus Gyroidella stood out as the only taxon significantly and positively correlated with coarse sand while being negatively correlated with both fine and very fine sand, marking it as a specialist of high-energy, coarse-grained settings.

Canonical correspondence analysis reinforced these findings in striking visual terms. The first two ordination axes together accounted for 90.9 percent of the constrained species-environment variance, with Cox’s Bazar samples aligning strongly with medium sands and Sundarbans samples clustering consistently with finer grains. Species vectors in the biplot closely tracked the sediment gradients, and the analysis independently reproduced the genus-substrate affinities identified by the correlation analysis. Finally, a two-way PERMANOVA based on Bray-Curtis dissimilarity and 9999 permutations confirmed that both location and sediment category significantly structured the assemblages, with p-values of 0.0001 and 0.0031 respectively. Crucially, the interaction term was not significant, meaning that sediment grain size exerts the same directional effect on community composition regardless of which of the three coastal environments is considered.

Why should grain size matter so much to organisms barely visible to the naked eye? The authors point to an indirect but well-supported mechanism: fine-grained sediments have greater surface area and adsorption capacity and are typically associated with higher organic matter concentrations, which in turn can deplete oxygen in the uppermost sediment layer. Although organic matter and dissolved oxygen were not directly measured in this study, a limitation the authors openly acknowledge, the dominance of opportunistic and stress-tolerant taxa such as Ammonia, Elphidium, and agglutinated forms in the Sundarbans is consistent with assemblages known from organically enriched coastal settings worldwide. The reduced diversity there likely reflects these stressed conditions, where a few tolerant species outcompete the rest.

The assemblages also encode hydrodynamic history. At Inani beach in Cox’s Bazar, the researchers found transported planktonic foraminifera, including Globigerina bulloides and Hantkenina alabamensis, species that live in the open water column and could only arrive on the beach through active wave and tidal transport. Their presence signals a well-connected, open-marine setting characteristic of the inner to middle neritic zone, rather than evidence of local productivity. In Kuakata, the combination of Elphidium, Cribroelphidium, Ammonia, and various miliolids, together with the complete absence of planktonic species, points to estuarine to very inner neritic conditions. In the Sundarbans, salt-marsh agglutinated foraminifera such as Trochammina inflata, Miliammina, Haplophragmoides, and Entzia were confined to areas under strong freshwater influence, echoing ecological patterns reported from karstic coasts of the Adriatic.

Beyond its scientific interest, the study carries practical weight for a coastline facing mounting natural and anthropogenic pressures, from sea-level rise to pollution and sediment starvation. By establishing sediment texture as a tractable, measurable, and directly comparable control on foraminiferal diversity, the work provides a framework for using these microfossils as bioindicators in environmental assessment. The authors recommend that future investigations incorporate direct measurements of organic matter, dissolved oxygen, salinity, and sediment mineralogy to refine the picture. For now, the message from the northern Bay of Bengal is clear: the grains on the seafloor, and the energy that arranges them, set the rules by which life at the smallest scale thrives or struggles, and reading those grains through the lens of foraminifera offers a powerful window into the health of coastal ecosystems under change.

Subject of Research: Relationship between sediment grain size and benthic foraminiferal diversity across contrasting coastal environments of Bangladesh

Article Title: Spatial patterns of benthic foraminiferal diversity in relation to sediment type across contrasting coastal environments of Bangladesh

Article References: Al Kabir, M. S., Saha, S. K., Tushar, M. A. N., & Maria, R. S. (2026). Spatial patterns of benthic foraminiferal diversity in relation to sediment type across contrasting coastal environments of Bangladesh. Discover Geoscience, 4(1), Article 344. https://doi.org/10.1007/s44288-026-00712-1

Image Credits: AI Generated

DOI: 10.1007/s44288-026-00712-1

Keywords: benthic foraminifera, sediment grain size, Bay of Bengal, Bangladesh coast, Sundarbans, Cox's Bazar, Kuakata, PERMANOVA, canonical correspondence analysis, coastal biomonitoring, microfossils, Ganges-Brahmaputra-Meghna delta

The research, led by Nuril Kusumawardani Soeprapto Putri of Institut Teknologi Sepuluh Nopember and Bina Nusantara University, together with Umi Laili Yuhana, Daniel Oranova Siahaan, and Istiani, addresses a problem familiar to anyone who has worked with large-scale assessment data. Universities routinely collect performance measures at admission, but these numbers are rarely translated into actionable profiles that instructors can use. The authors argue that before artificial intelligence can meaningfully stratify learners, the analytics pipeline itself must be reproducible, leakage-free, and honest about what a statistical cluster does and does not represent. Their framework is explicitly designed so that clusters are treated as operational descriptions of performance patterns, not as naturally discrete or rank-ordered psychological categories.

At the foundation of the approach lies multidimensional item response theory, a psychometric technique that estimates latent abilities from how students answer individual test items rather than relying on raw totals. The researchers extracted ability estimates in three domains: logical reasoning, abstraction, and verbal reasoning. This multidimensional representation matters because two students can arrive at the same overall score through very different routes. One might excel in verbal tasks while struggling with logic, while another shows the reverse pattern. A single aggregate number collapses those differences; a three-dimensional ability vector preserves them and makes richer profiling possible.

To guard against data leakage, one of the most insidious pitfalls in applied machine learning, the cohort was split into a training set of 29,449 students and a held-out test set of 7,364 students. Every transformation, every cluster boundary, and every profile label was derived exclusively from the training partition and then transferred, unchanged, to the test partition. This discipline ensures that the reported performance reflects how the system would behave on genuinely new students rather than on data it has already seen. Many published clustering studies skip this step entirely, fitting and evaluating on the same data and thereby inflating apparent quality. Here, the split is the backbone of the entire validation strategy.

The first level of the framework applied k-means clustering with the number of clusters fixed at four, a choice justified by a composite ranking of three standard internal validation indices: the Silhouette score, which measures how well each point fits its own cluster compared to neighbors; the Davies-Bouldin index, which rewards compact, well-separated clusters; and the Calinski-Harabasz index, which compares between-cluster to within-cluster dispersion. Using multiple indices rather than a single criterion reduces the risk of selecting a cluster count that flatters one metric but fails on others. The resulting four profiles were characterized not by how much ability students had, but by the shape of their ability centroids: a uniformly lower profile, a uniformly higher profile, a logic-dominant profile, and a verbal-abstraction dominant profile.

One of the most striking findings comes from the relationship between those two middle profiles. The logic-dominant and verbal-abstraction dominant groups crossed each other across the three ability dimensions, meaning that neither could be described as simply better than the other. If cognitive performance lay along a single ordinal continuum from low to high, such crossing would be impossible. Its presence demonstrates, in the study’s own data, that student cognition cannot be compressed onto one severity scale. Students are not merely higher or lower performers; they exhibit distinct patterns of relative strengths, and those patterns carry information that intervention designers can exploit.

The second level of the pipeline drilled into the uniformly lower profile, subdividing it into four domain-specific sub-profiles. This stratification is arguably the study’s most practically valuable contribution. Students in the lowest overall group are not interchangeable; some struggle more with logic, others with abstraction or verbal reasoning. By separating these sub-groups, the framework gives educators a way to tailor remediation, for example by directing logic-focused support to one sub-profile and verbal or comprehension-focused support to another, rather than applying a generic intervention to everyone below a cutoff. Differentiated support of this kind has long been an aspiration in higher education; this work offers a scalable computational route toward it.

The statistical evidence for the clustering’s strength was substantial. A PERMANOVA test, a non-parametric analysis of variance suited to multivariate distance data, yielded a pseudo-F statistic of 10,798 and an R-squared of 0.524, indicating that the profile assignment accounted for more than half of the multivariate variance in the ability estimates. Among the three cognitive domains, logical reasoning contributed most strongly to differentiating the profiles, with an effect size of eta-squared equal to 0.81. These are large effects by any convention in the behavioral sciences, and they indicate that the four-way partition captures genuine structure in the data rather than an arbitrary slicing of a homogeneous cloud.

The authors are notably careful about what the final validation step proves. They used hierarchical label recovery to reconstruct profile assignments on the held-out test data, achieving approximately 98 percent accuracy, which demonstrates that the assignment rule transfers reliably to unseen students. Yet they also report that a simple baseline model achieved comparable performance, and they interpret this honestly: the result demonstrates reproducibility of the assignment rule under leakage control, not psychological validation of the clusters as true categories. This distinction matters. The clusters are useful operational summaries, but the framework does not claim that four cognitive types exist in nature. That kind of epistemic restraint is rare and welcome in a field often tempted to overinterpret unsupervised structure.

Broader implications extend beyond the specific dataset. The study received institutional support from the INNOGEN Lab of Bina Nusantara University and was conducted as a retrospective analysis of fully anonymized secondary data, with ethical clearance from the university’s research ethics committee. Because the pipeline is computationally scalable and the labels are interpretable, the approach could in principle be adopted by institutions that already administer multidimensional admission or placement assessments. The framework offers an interpretable and efficient way to convert raw assessment data into profiles that support differentiated educational interventions, while its emphasis on held-out validation and modest claims sets a methodological example for the growing field of learning analytics. As universities increasingly turn to data-driven tools to support struggling students, this work suggests that the most important algorithmic ingredient may be rigor: get the measurement right, split the data honestly, and never mistake a convenient cluster for a psychological fact.

Subject of Research: A reproducible two-level machine learning framework for cognitive profiling and stratification of low-performing students in higher education

Article Title: A reproducible two level clustering framework for profiling cognitive performance and stratifying the lowest performing group in higher education

Article References: Putri, N. K. S., Yuhana, U. L., Siahaan, D. O., & Istiani (2026). A reproducible two level clustering framework for profiling cognitive performance and stratifying the lowest performing group in higher education. Discover Psychology. https://doi.org/10.1007/s44202-026-00888-0

Image Credits: AI Generated

DOI: 10.1007/s44202-026-00888-0

Keywords: cognitive profiling, item response theory, k-means clustering, cluster validation, reproducibility, higher education, learning analytics, psychometrics, data leakage, PERMANOVA, educational intervention, machine learning