The research, conducted by Noora Poikela, Rhonda R. Snook, Jonna Kulmuni, and Michael G. Ritchie, focused on two closely related species of the virilis group of Drosophila. Drosophila flavomontana is comparatively sensitive to heat, while its relative Drosophila montana tolerates high temperatures far better. The two species overlap in parts of their natural ranges, yet they remain distinct, maintained by a web of genetic incompatibilities that biologists call barrier loci. These are stretches of the genome where alleles from one species function poorly, or not at all, in the genetic background of the other, so natural selection removes them whenever hybridization occurs.

Barrier loci pose a genuine puzzle for adaptation. If the very regions that resist gene flow sit close to genes that could confer useful traits, then a species in crisis may be locked out of its best source of rescue. The team asked whether an intense, sustained selective pressure such as heat stress could crack these barriers open and let adaptive alleles through. They also asked a subtler question about genome architecture: chromosomal inversions, in which a segment of a chromosome is flipped relative to the ancestral arrangement, suppress recombination. Inversions can therefore bind adaptive genes and barrier genes together into inherited blocks, either promoting their joint transfer or freezing the boundary in place.

To test these ideas, the researchers ran what is known as a hybridize, evolve, and re-sequence experiment. They crossed the heat-sensitive D. flavomontana with the more tolerant D. montana to create admixed populations, then split the offspring into experimental lines. Heat-selection lines were exposed to transient but punishing heat for three consecutive generations, while parallel control lines were maintained under benign conditions. Afterward, the team sequenced the genomes of every line and tracked where DNA from D. montana had persisted. In parallel, they performed a demographically explicit genome scan of natural populations of both species to map the barriers to gene flow that have accumulated over evolutionary time, giving them a long-term benchmark against which to judge the experiment.

The results were striking. Lines that had experienced heat stress showed roughly twice as much introgressed DNA as the control lines. Just as importantly, the heat-selected lines retained higher male fertility under heat stress than both the control lines and the pure, heat-sensitive D. flavomontana parent. Male fertility is a critical measure in these flies, because hybrid male sterility is one of the classic barriers that keeps closely related Drosophila species separate. The fact that heat selection preserved male reproductive function while admitting foreign genes indicates that the introgressed material was genuinely adaptive rather than merely neutral hitchhikers surviving by chance.

Genome-wide mapping revealed where this transfer happened. The bulk of the extra introgression in heat-selected lines sat in colinear autosomal regions, stretches of the chromosomes whose gene order is the same in both species and where recombination proceeds normally. The amount of introgression in these regions correlated with the degree of improvement in heat tolerance, tying the borrowed DNA directly to the adaptive phenotype. This pattern supports the idea that when selection is strong enough, it can sweep adaptive alleles across species boundaries through ordinary, recombining parts of the genome, provided the fitness benefit outweighs the cost of any linked incompatibilities.

Inversions told a more nuanced story. Some introgression did occur in inverted autosomal regions, but only in the heat-selection lines and never in the controls. That contrast suggests inversions are not absolute fortresses: under intense selection, the tight linkage they create can occasionally carry adaptive combinations across, but the process is far less efficient than in colinear DNA. The clearest exception was the X chromosome. Carrying three overlapping inversions, it resisted introgression in both heat-selected and control lines alike. The X chromosome therefore maintained a robust species barrier even under conditions that opened the autosomes, consistent with its well-known role in hybrid incompatibility and sterility in flies.

The comparison between experimental and natural barriers added an evolutionary timescale to the picture. Genetic barriers that emerged in the control lines, where heat selection was absent, overlapped significantly with the long-term barriers identified in natural populations by the genome scan. In other words, left to their own devices over a few generations, admixed fly genomes retrace the same boundaries that separate the species in the wild. In the heat-selection lines, that overlap was noticeably weaker, showing that strong thermal selection had actively eroded parts of the ancient barrier architecture and allowed a different, more permeable genomic outcome.

Taken together, the study delivers a message with real urgency for conservation and evolutionary biology. Adaptive introgression is not just a theoretical curiosity; it can operate on the timescale of a handful of generations when selection is severe, and it can deliver measurable gains in a trait as directly tied to survival as heat tolerance. For small populations facing local extinction as climates warm, related species nearby may represent living archives of pre-tested adaptive alleles. The work also carries a caution: not all of the genome is equally available for borrowing. Regions dense with incompatibilities, and especially heavily inverted chromosomes such as the X, may continue to resist gene flow no matter how hot it gets. Understanding which parts of a genome are permeable, and which remain sealed, will be essential for predicting which species can adapt by hybridization and which must rely on standing variation or new mutation alone. The fruit flies of northern latitudes, it turns out, have been running this experiment for millennia, and now science has caught a glimpse of the answer.

Subject of Research: Adaptive introgression and heat tolerance evolution between two Drosophila species

Article Title: Adaptive introgression between two Drosophila species enhances heat tolerance despite barriers to gene flow

Article References: Adaptive introgression between two Drosophila species enhances heat tolerance despite barriers to gene flow. (n.d.). https://doi.org/10.1371/journal.pbio.3004028

Image Credits: AI Generated

DOI: 10.1371/journal.pbio.3004028

Keywords: adaptive introgression, Drosophila, heat tolerance, climate change, barrier loci, chromosomal inversions, hybridization, genomics, experimental evolution, X chromosome, speciation, PLOS Biology

The team, led by Nico Appold, Timon Citak, Auguste Palm and Jona Kayser, built a microbial therapy-mimicry assay around the budding yeast Saccharomyces cerevisiae. Populations were initiated from single, genetically tailored cells placed in a high-throughput arrayed colony assay, where each founder cell expanded into a compact, range-expanding colony. To generate a continual supply of trackable resistant mutants, the researchers engineered a stochastic, irreversible synthetic mutation system based on a hierarchical dual-recombinase switch. In the unmutated state, cells are thermosensitive: they proliferate at 30 degrees Celsius but arrest at 35 degrees. Upon switching, cells become thermoresistant and simultaneously change fluorescence colour, allowing resistant lineages to be identified and followed throughout colony growth. Treatment itself is delivered through programmable temperature shifts, giving the experimenters precise control over therapy onset, duration and pause.

Combining this assay with multi-day time-lapse fluorescence microscopy and automated temperature control, the researchers recorded population histories for up to 160 hours, spanning more than 100 generations. A custom image-analysis pipeline using machine learning-based segmentation extracted colony- and clone-level dynamics from the raw imaging data. Radial kymographs, which trace population structure along the direction of expansion, revealed two qualitatively distinct fates for resistant clones. Some escaped confinement as continually growing domes at the colony edge, but most remained trapped within the population bulk. This baseline phenomenology set the stage for the central question: what determines whether a bulk-born mutant escapes or stays confined?

Under continuous therapy, the answer proved to be strikingly spatial. Although the temperature shift was spatially homogeneous, mutant expansion showed pronounced spatiotemporal structure. The time at which a resistant mutant was first detected correlated strongly with its distance from the colony front, indicating that mutant growth was gated by a spatially varying growth potential within the colony. Individual mutant trajectories crept towards the colony edge until permanent escape occurred, and mutants originating further inward sometimes remained confined when a more forward lineage occupied the advancing front. At the colony level, this created a finite time window after therapy onset during which no mutant had yet escaped; as therapy continued, the cumulative number of escaped mutants rose steadily.

The existence of this delay suggested a therapeutic lever. If treatment were paused before any mutant reached the edge, the resumption of sensitive growth might re-establish a competitive barrier at the front. The researchers tested this with a 14-hour therapy pulse followed by a treatment holiday. Strikingly, many mutants that had expanded during the pulse stalled and remained confined within the colony bulk, with re-confinement often preceding the full recovery of sensitive expansion. Systematically varying pulse duration showed that shortening treatment strongly enhanced confinement: for pulses shorter than six hours, resistance escape was almost completely suppressed. The maximum safe pulse length proved more restrictive than continuous-therapy trajectories alone would suggest, because mutants continue advancing during the post-pulse regrowth lag.

To explain these observations mechanistically, the team constructed a stochastic simulation framework, a mechanistically minimal digital twin in which nutrient dynamics are directly accessible. Space is discretized into a two-dimensional grid of demes containing sensitive and resistant cells that proliferate at nutrient-dependent rates and disperse through growth-driven spreading. Competition is mediated solely through a shared, freely diffusing and consumable nutrient field, with no explicit local carrying capacity. Therapy is implemented as a modulation of sensitive growth that captures experimentally observed response kinetics, including on-delays and post-treatment lags. Despite its minimal structure, the model qualitatively reproduced the key experimental dynamics: bulk-born mutants released under therapy, and re-confinement when treatment pauses allowed sensitive growth to rebound.

The digital twin revealed the underlying resource logic. In a range-expanding population, growth and nutrient consumption generate a nutrient gradient along the expansion direction, and the advancing front continuously couples the colony edge to the high-nutrient side of this gradient. Under continuous therapy, suppressed sensitive growth reduces nutrient consumption, allowing nutrients to penetrate inward and expanding the growth-permitting region into the bulk. This releases previously confined resistant mutants and ultimately enables them to reach and escape at the colony edge. Pausing therapy reverses the process: renewed sensitive growth restores nutrient depletion in the bulk and reforms a competitive barrier of sensitive cells at the front, restoring spatial confinement.

With this mechanism in hand, the researchers scanned a large space of intermittent schedules, varying the durations of therapy-on and therapy-off intervals. Performance was quantified by time to progression, the time required for colony area to exceed a predefined threshold. Across schedule space, time to progression exhibited a pronounced ridge of optimal schedules separating two modes of failure: sensitive-dominated progression under too little treatment and resistant-dominated progression under too much. A sharp switch in the resistant fraction occurred near the optimum, which the authors describe as a dynamic phase-transition-like boundary in schedule space. Beyond the critical pulse length, therapy depresses front growth strongly enough that recovery during each pause is incomplete, producing a ratchet-like drift that progressively weakens confinement until resistant clones escape.

Guided by the simulations, the team tested the predicted sweet spot experimentally using pulse durations of four, 6.5 and nine hours with 18-hour pauses. The experiments confirmed a local optimum: time to progression peaked at 6.5-hour pulses, which significantly outperformed continuous therapy and both adjacent schedules, and the resistant fraction rose sharply near the transition. The 9-hour schedule reproduced the model-predicted cumulative decline in sensitive growth and surge in mutant escapes. As a proof of principle, the researchers also ran an adaptive therapy experiment in which treatment intervals were adjusted in real time based on measured growth rebound rather than fixed intervals. This closed-loop strategy slowed overall growth more than any fixed schedule, typically delaying progression beyond the experimental runtime, although resistant-population dynamics were more variable.

The study establishes resource-mediated spatial confinement as a central organizing principle of resistance evolution and offers a mechanistic foundation for spatially informed, evolution-based therapies. The authors suggest that growth rebound could serve as a clinically accessible measure of confinement potential, and that poorly vascularized tumours with steep resource gradients may be ideal candidates for such approaches, potentially enhanced by angiogenesis-targeting drugs. The findings may also explain why recent clinical trials of intermittent therapy failed to improve progression-free survival: the therapeutic sweet spot is narrow, and hitting it requires accounting for spatial resource dynamics that mean-field models ignore. The researchers caution that their yeast assay is deliberately reductionist, and that testing confinement-and-release dynamics in three-dimensional cancer organoids and matrix-producing biofilms will be essential before drawing clinical conclusions. Still, the message is clear: whether therapy fails may hinge on whether treatment shifts the resource gradient long enough for bulk-born mutants to permanently escape confinement.

Subject of Research: Spatial resource dynamics governing the emergence and escape of therapy-resistant mutants in structured microbial populations

Article Title: A microbial therapy-mimicry assay shows how spatial resource dynamics control resistance escape

Article References: A microbial therapy-mimicry assay shows how spatial resource dynamics control resistance escape. (n.d.). https://doi.org/10.1038/s41559-026-03178-z

Image Credits: AI Generated

DOI: 10.1038/s41559-026-03178-z

Keywords: therapy resistance, spatial evolution, nutrient gradients, intermittent therapy, adaptive therapy, yeast colonies, digital twin, biofilms, cancer evolution, experimental evolution, phase transition, range expansion

The team’s central contribution is a matrix representation of genotypic selection. Instead of writing out separate equations for every possible genotype, the authors encode the fitness consequences of every allele pairing in a single symmetric matrix, which they call the fitness-effect matrix. Each element of this matrix quantifies how much a particular genotype departs from a neutral baseline: a value of zero means the genotype enjoys no selective advantage or disadvantage, while positive or negative values signal fitness gains or losses. Because the matrix is symmetric, with the fitness of the heterozygote carrying allele i and allele j identical to that of the reverse pairing, the entire selection regime at a locus with n alleles is compressed into n(n+1)/2 numbers. This compact encoding exploits a deep mathematical symmetry: the same matrix structures that govern deterministic selection also describe the variance of random genetic drift, allowing the two great engines of evolutionary change to be handled within one coherent formalism.

One of the framework’s most appealing features is its flexibility. The authors show that the matrix representation accommodates an impressive range of selection regimes without modification. Additive selection, where each allele contributes independently to fitness; multiplicative selection, where effects combine like compound interest; frequency-dependent selection, where the advantage of an allele shifts as its prevalence changes; temporally varying selection, where the environment flips fitness rankings across generations; and heterozygote advantage, the famous mechanism that maintains sickle-cell anaemia at high frequencies in malaria-endemic regions, all fall naturally within the same mathematical structure. This generality matters because heterozygote advantage and other multiallelic interactions are precisely the cases where two-allele approximations break down most dramatically, producing stable polymorphisms and oscillatory dynamics that biallelic models misinterpret or miss entirely.

The biological motivation for the work is easy to appreciate. The major histocompatibility complex, the gene cluster that underpins the vertebrate immune system’s ability to recognise pathogens, is among the most polymorphic regions of the human genome, with thousands of alleles described. The ABO blood group system, with its three principal alleles and their combinatorial genotypes, shapes transfusion medicine and has been linked to susceptibility to various diseases. Genes underlying monogenic disorders likewise often harbour multiple disease variants. At each of these loci, an individual carries only two alleles, but the population as a whole can carry many, generating a rich spectrum of genotypes whose relative fitnesses determine how variation is maintained or lost. Understanding selection in these settings requires tools that treat the full genotype matrix, not a collapsed two-allele shadow of it.

Describing selection is only half the problem; the other half is measuring it. Here the study makes a second, arguably more practical, contribution. The authors derive a procedure for estimating the fitness-effect matrix from allele frequency trajectories, the kind of data now routinely generated by experimental evolution and by ancient DNA and ecological time-series sampling. The logic is elegant: given a candidate matrix, one can predict how allele frequencies should change from one generation to the next, and then compare those predictions with the observed trajectory. The researchers formalise this comparison in a cost function, weighted by the Shahshahani metric, a geometric measure of distance between allele frequency vectors that arises naturally from the variance structure of genetic drift. When the candidate matrix equals the true one, the cost vanishes; otherwise, discrepancies between predicted and observed frequencies inflate the cost, and a numerical optimisation procedure adjusts the matrix until the best fit is found.

Several technical refinements make the estimation procedure robust in practice. Because relative fitnesses are only defined up to a multiplicative constant, the authors fix one genotype as a reference with a vanishing fitness-effect, removing a redundancy that would otherwise make the problem ill-posed. They also impose symmetry on the matrix throughout the optimisation, ensuring that the heterozygote fitnesses remain consistent regardless of allele ordering. To handle the practical realities of noisy, sparse data, the cost function includes a small constant that prevents infinite weightings when an allele’s frequency dips toward zero, and a ridge penalty that shrinks estimated fitness effects toward zero, suppressing the wild values that small samples can otherwise produce. The authors report that their results are largely insensitive to the first of these tuning parameters, while the optimal strength of the ridge penalty decreases with population size, a pattern they verified systematically across simulations.

The framework also handles an awkward gap that plagues real datasets: sampling. Experimental populations are rarely censused every single generation, and natural populations even less so. The authors extend their method to trajectories in which allele frequencies are known only at irregular intervals. In this case, the prediction is propagated forward through all the unsampled generations, with each intermediate step feeding into the next, until the model reaches the next sampling point. The cost is then computed only at the sampled generations. This means the method can extract fitness information from patchy time-series, provided the underlying dynamics remain governed by the same selection regime across the gaps, an assumption the authors are careful to state.

To demonstrate the approach on real data, the team applied their estimation procedure to time-series data from an experimental yeast evolution study. Yeast populations, with their rapid generations and tractable haplotype tracking, provide an ideal proving ground for microevolutionary inference. The analysis illustrated that multiallelic fitness interactions, including signatures consistent with heterozygote advantage, could be characterised directly from haplotype frequency data, without the need to reduce the system to pairwise biallelic comparisons. This demonstration is significant because it shows the method working end to end: from raw frequency trajectories, through the optimisation machinery, to a fitted matrix of genotype-specific fitness effects that tells a coherent evolutionary story.

The broader implications reach across several fields. For experimental evolution, the framework offers a principled way to quantify fitness landscapes at highly polymorphic loci, where epistatic interactions between alleles within a gene can be mapped and compared across environments. For conservation and evolutionary biology, it provides tools to detect balancing selection at immune genes and other multiallelic systems in natural populations, using the increasingly affordable time-series data generated by ecological genomics programmes. For medical genetics, a better handle on multiallelic selection could sharpen understanding of how disease-associated variation persists in populations despite its costs. The work, funded in part by the European Research Council under the Horizon 2020 programme, is published open access, and the authors have released the full mathematical derivations, including the stochastic dynamics under a Wright-Fisher model and a diffusion approximation, in the article and its supplementary material.

None of this means the two-allele models that dominate the literature are wrong; they remain excellent approximations when a locus genuinely harbours few variants, and their simplicity has powered a century of insight. But as genomic data grows richer and time-series sampling becomes routine, the loci that matter most, from the immune genes that fight our infections to the variants that shape our blood and our diseases, increasingly demand a mathematics that matches their diversity. The fitness-effect matrix of Vellnow, Gossmann and Waxman provides exactly that: a compact, flexible and estimable description of selection that finally looks beyond two alleles and sees the full, tangled richness of genetic variation as evolution actually experiences it.

Subject of Research: A matrix framework for modelling and estimating multiallelic genotypic selection from allele frequency time-series data

Article Title: Beyond two alleles: multiallelic genotypic selection and its estimation from time-series data

Article References: Vellnow, N., Gossmann, T. I., & Waxman, D. (2026). Beyond two alleles: multiallelic genotypic selection and its estimation from time-series data. BMC Genomics. https://doi.org/10.1186/s12864-026-13411-5

Image Credits: AI Generated

DOI: 10.1186/s12864-026-13411-5

Keywords: population genetics, multiallelic loci, genotypic selection, time-series data, heterozygote advantage, genetic drift, fitness estimation, Wright-Fisher model, experimental evolution, yeast, BMC Genomics, balancing selection

The research team set out to compare five competing approaches for detecting polygenic adaptation from time-series data on allele frequency changes. Three of these were single methods: the Fisher’s Exact Test, a venerable statistical workhorse that compares allele counts between sampling time points; a Naive Bayesian Classifier, a probabilistic machine learning model that assigns observations to categories based on learned feature distributions; and a One-Class Support Vector Machine, an anomaly-detection algorithm that learns the shape of ‘normal’ neutral evolution and flags deviations from it. The remaining two approaches were hybrids, combining the machine learning models with the classical test to produce OCSVM-FET and NBC-FET. The logic behind the hybrids is intuitive but powerful. The Fisher’s Exact Test is sensitive to statistically significant allele frequency shifts at individual loci, while the machine learning models can integrate patterns across features and recognize the subtle, correlated structure that polygenic selection imposes on many loci at once. By combining the two kinds of evidence, the hybrid methods can filter out noise that would fool either approach alone.

A crucial challenge in this field is that real genomes under known selection pressures are hard to come by. In natural populations, researchers rarely know which loci are actually under selection, how strong the selection is, or when it began, which makes it nearly impossible to judge whether a detection method is working. The authors therefore turned to simulated data generated within an evolve-and-resequence framework, a design in which researchers can specify the exact evolutionary scenario, including the number of loci under selection, the strength of selection, the number of generations, and the timing of genomic sampling, and then evaluate how well each method recovers the truth. The simulations were parameterized using empirical Pool-Seq data from the non-biting midge Chironomus riparius, an organism well suited to experimental evolution studies, which grounds the synthetic scenarios in biologically realistic levels of genetic variation and demographic noise.

Pool-Seq itself deserves a brief technical explanation, because it shapes both the strengths and the limitations of the data. Rather than sequencing individual organisms separately, Pool-Seq sequences an entire population sample together, estimating allele frequencies from the pooled reads. This makes time-series sampling affordable: researchers can sequence a population at generation zero, generation ten, generation twenty, and so on, tracking how allele frequencies shift over time. For detecting selection, these temporal trajectories are gold, because neutral drift produces random walk-like fluctuations while selection produces directional, consistent shifts. However, Pool-Seq introduces its own measurement uncertainty, since read depth and pooling noise mean that estimated frequencies are never exact. Any detection method must therefore distinguish genuine selective signals from both stochastic drift and sampling error, a task that becomes exponentially harder when the per-locus effects of selection are as small as those expected under polygenic adaptation.

The evaluation across scenarios varying in generation number, selection strength, and the number of selected loci produced a clear ranking. The combined OCSVM-FET approach achieved the lowest false positive rate, the highest area under the receiver operating characteristic curve, and high overall accuracy, consistently outperforming the standalone machine learning models, the standalone Fisher’s Exact Test, and the alternative NBC-FET hybrid. In practical terms, this means that when the OCSVM-FET method flags a locus or a pattern as being under selection, that flag is more likely to be real, and the method captures a larger fraction of the true targets, than any of the competing approaches tested. For experimental evolutionists planning expensive sequencing campaigns, this reduction in false positives is not a statistical nicety; it directly determines which candidate loci get followed up and how much laboratory validation is required.

Perhaps the most biologically interesting result concerns timing. The performance of the winning method peaked during what the authors call the ‘late dynamic phase’ of adaptation, the window after selection has begun to shift allele frequencies but before the favored variants have reached fixation. This makes deep evolutionary sense. Early in adaptation, allele frequency changes are so small that they drown in drift and sampling noise, and even the best methods have little to work with. Late in adaptation, once beneficial alleles have fixed, the variation that methods rely on has been erased, and the signal vanishes again. The late dynamic phase is the sweet spot in between, where selection has produced measurable, directional movement at many loci while sufficient polymorphism remains for statistical tests and machine learning models to detect structure. The finding carries an immediate practical message for experimental design: temporal sampling schemes should be tuned to capture this phase, because sequencing too early or too late can render even a superior method blind.

The study also delivers a sobering lesson about parameter tuning, which the authors describe as critical and as a matter of balancing biological assumptions against methodological rigor. Machine learning models are not plug-and-play devices; the One-Class Support Vector Machine, for instance, requires choices about kernel functions, nu parameters, and the feature representation of the data, and each of these choices encodes implicit assumptions about what neutral evolution looks like and how selection deviates from it. Set the parameters too loosely and the model flags noise as signal; set them too tightly and genuine adaptation slips through. The authors emphasize that these choices must be informed by the biology of the system, including realistic effective population sizes, mutation rates, and the expected architecture of the trait under selection, rather than by generic defaults. This is a point that resonates well beyond this particular study, as machine learning methods proliferate across genomics faster than the standards for validating and tuning them.

The authors are careful about scope, and that caution matters for how the results should be used. The benchmarking was intentionally designed and validated for evolve-and-resequence experimental contexts, where defined selection pressures and temporal sampling are feasible, and the authors note that applicability to certain natural experiments might be possible but remains to be demonstrated. Natural populations present complications that the controlled simulations do not fully capture: population structure, admixture, changing demography, and unknown and heterogeneous selection regimes can all mimic or mask polygenic signals. Extending the framework to those settings is explicitly flagged as an important direction for future work. Until then, the strongest claims apply to laboratory and semi-natural experimental evolution studies, which are, fortunately, a growing and increasingly important part of evolutionary genomics, particularly for questions about rapid adaptation to environmental change.

The broader significance of this work lies in what it suggests about the future of method development in evolutionary biology. For decades, the field has relied on a relatively small toolkit of statistical tests, each with well-understood assumptions and limitations. Machine learning offers a way to move beyond single-locus tests toward classifiers that integrate many weak signals into a coherent decision, but the study shows that the most effective strategy is not replacement but combination. The classical test contributes a well-calibrated measure of per-locus evidence; the machine learning model contributes pattern recognition across loci and features; and the fusion of the two achieves better discrimination than either alone. This hybrid philosophy, in which old and new methods are treated as complementary sources of evidence rather than rivals, may prove to be a template for tackling other detection problems in genomics where signals are subtle, distributed, and easily drowned by noise.

For researchers working on experimental evolution, the study provides an actionable benchmark: when designing evolve-and-resequence experiments aimed at dissecting the genomic basis of quantitative traits, the OCSVM-FET approach currently offers the best validated performance for detecting polygenic adaptation from Pool-Seq time series, provided that sampling captures the late dynamic phase and that parameters are tuned to the biology of the study organism. For the wider community, the message is that the era of polygenic adaptation detection, long hampered by the sheer subtlety of the signal, may be entering a more productive phase, one in which carefully benchmarked hybrid methods, grounded in realistic simulations and honest about their limits, replace the optimistic one-size-fits-all tests of the past. As climate change accelerates the pace of adaptation in natural and experimental populations alike, the ability to see those small, distributed shifts in the genome will only grow in importance, and this study marks a meaningful step toward that goal.

Subject of Research: Detection of polygenic adaptation using machine learning and statistical methods on simulated evolve-and-resequence genomic data

Article Title: Enhancing detection of polygenic adaptation: a comparative study of machine learning and statistical approaches using simulated evolve-and-resequence data

Article References: Caliendo, C., Gerber, S., & Pfenninger, M. (2026). Enhancing detection of polygenic adaptation: a comparative study of machine learning and statistical approaches using simulated evolve-and-resequence data. BMC Genomics, 27(1), Article 740. https://doi.org/10.1186/s12864-026-13163-2

Image Credits: AI Generated

DOI: 10.1186/s12864-026-13163-2

Keywords: polygenic adaptation, machine learning, population genomics, evolve-and-resequence, Pool-Seq, One-Class Support Vector Machine, Fisher's Exact Test, allele frequency, experimental evolution, BMC Genomics, false positive rate, time-series genomics

Loss of heterozygosity, or LOH, is a familiar concept in genetics. In a diploid organism, each gene normally exists in two copies, one inherited from each parent. Errors during mitosis, the process by which a cell duplicates and divides, can eliminate one of those copies, either through recombination, chromosome loss or segmental deletion. Previous work had established that mitotic errors alter the allelic composition of a gene roughly a thousand times more often than replication errors produce point mutations. This means that when selection favors the loss of a gene’s function, it will frequently be achieved by discarding the chromosome or chromosomal segment that carries it, rather than by mutating the gene itself. Aneuploidy, the state of having an abnormal number of chromosomes, is found in up to a quarter of newly isolated wild strains of Saccharomyces cerevisiae, a striking testament to how common these events are.

What remained unclear was whether complex rearrangements, those affecting multiple chromosomes at once, accumulate gradually through a series of independent steps or emerge suddenly in coordinated bursts. To find out, the team built an experimental system in which genome rearrangements could be generated de novo and their frequency, triggers, architecture and stability all measured. They compiled a list of 454 yeast genes whose deletion had been shown, in earlier screens using two well-established but distinct methods, one detecting gross chromosomal rearrangements and one detecting chromosome instability, to increase LOH. After accounting for strains that could not be manipulated successfully, 392 diploid deletion strains remained, each carrying a homozygous deletion of a different gene together with two heterozygous marker loci: CAN1 on the left arm of chromosome V and ade2::URA3 in the middle of the right arm of chromosome XV.

The researchers propagated 20 replicate mutation-accumulation cultures of each strain and plated cells onto selective media containing canavanine or 5-fluoroorotic acid, which only cells that had lost the functional allele at the corresponding marker could colonize. These assays yielded robust, strain-specific estimates of the single LOH rate per cell division, with 95 percent confidence limits typically spanning only a narrow range around the central value. A substantial majority of the deletion strains, 68 percent at the CAN1 locus and 82 percent at the URA3 locus, showed higher single LOH rates than the wild-type control, confirming that the mutations compiled from earlier qualitative screens indeed destabilize the genome in quantitative assays. The team then scored how frequently double mutants, resistant to both drugs simultaneously, appeared. Deletion strains varied enormously, and the most unstable strains under double selection were also among the most unstable under single selection, with a Spearman rank correlation of 0.61.

The critical question was whether these double events could be explained as two independent single events occurring in sequence. Because the experimental data alone could not distinguish one-step from two-step origins, the team turned to computer simulations. Using the empirically measured single LOH rates, they modeled mutation accumulation across colony growth, allowing cells to become single mutants at their measured rates and double mutants either sequentially or coincidentally at a rate equal to the product of the two single rates. Even under the deliberately conservative assumption that neither single nor double mutants paid any fitness penalty, an assumption that would maximize the survival and spread of sequentially derived double mutants, the simulated frequencies fell far short of what was observed. When fitness costs of 10 and 20 percent were applied, as is more realistic, the gap widened further. The observed double LOH frequencies exceeded the coincidental predictions by tens or even hundreds of times in the most unstable strains, and after subtracting the predicted sequential component, the residual coincidental events remained 26 times more frequent than expected.

What kinds of cellular defects produce such explosive instability? The answer emerged from Gene Ontology analysis of the deleted genes in the most unstable strains. Nearly half of the genes carry annotations related to cell cycle checkpoints, the surveillance mechanisms that halt progression through the division cycle when DNA is damaged or the mitotic spindle is improperly assembled. The checkpoint category as a whole is assigned to only 1.3 percent of all yeast genes, so its dominance among the strongest mutators is statistically striking. The remaining genes in the set are involved in DNA maintenance and repair, sister chromatid cohesion, and spindle formation, processes whose failure can generate damage so extensive that even intact checkpoints cannot contain it. The message is clear: genomes are most vulnerable to complex, multi-chromosome rearrangements when the checkpoints that police DNA integrity and spindle function fail to stop the cell cycle.

Direct examination of the double mutants themselves confirmed the picture. Flow cytometry of roughly 80 independent mutants from each of 43 highly unstable deletion strains revealed that DNA content was often substantially increased or decreased relative to the diploid ancestors, with the variance in DNA content among mutants significantly exceeding that among ancestral replicates. Whole-genome sequencing of mutants from 20 deletion strains went further. The dominant alteration was whole-chromosome aneuploidy, frequently affecting multiple chromosomes beyond the two targeted by selection, with many chromosomes showing non-integer coverage values that indicated mixed populations of cells carrying different chromosome counts. Averaged across all strains and chromosomes, deviations from diploidy appeared in nearly every mutant examined. Crucially, the chromosomes under direct selection, V and XV, showed no more than the average number of changes, evidence that selection at two sites was merely revealing systemic destabilization rather than focused local mutation. Heterozygous point mutations were rare, and the team found essentially no segmental copy number variants, no non-homologous recombination products and no convincing translocations or inversions in non-telomeric regions, despite deliberately validating that their pipelines could detect all such events.

Genetic tests of four particularly unstable strains, deleted for sem1, chl1, mre11 or rad27, provided independent confirmation. When double mutants from these strains were screened for retention of heterozygosity at two additional marker loci on chromosomes II and XII, the frequencies of additional losses were extraordinarily high. In the sem1 background, for example, 16 of 61 double mutants had also lost heterozygosity on chromosome II and 14 on chromosome XII, with 7 losing both, even though such losses normally occur at a rate of only one to a few per ten thousand cell divisions. Triple and quadruple LOH events appeared at frequencies far beyond anything a sequential model could accommodate, reinforcing the conclusion that short periods of hyper-instability, rather than a string of independent accidents, generate these genomes.

The final experiments addressed what happens to these dramatically rearranged genomes over evolutionary time. The team propagated 63 mutant cultures derived from 21 deletion strains for 200 generations in batch culture and re-measured their DNA content. The result was a nearly uniform contraction toward diploidy: strains that started far from two chromosome sets moved markedly closer, and the most divergent isolates showed the greatest fitness gains, growing faster after evolution than before. Aneuploidy, in other words, is typically a transient state, imposed as a cost of rapid adaptation and then pruned by natural selection as faster-growing re-diploidized variants outcompete their unstable relatives. The authors suggest that this ephemerality is exactly why such bursts are so hard to catch in nature, since selection rapidly erases the evidence of a genomic episode.

The findings carry implications well beyond yeast. In free-living microbes, aneuploidy usually slows growth and is quickly eliminated, but it may pay off under multiple stresses, such as exposure to several toxins at once. In the somatic cells of multicellular organisms, the calculus is different. These cells are programmed not to divide, so there is no immediate growth cost to genome restructuring, and a loosening of the brakes on cell division can incidentally expose recessive alleles, a mechanism central to cancer, where the loss of the remaining wild-type copy of a tumor suppressor gene is the classic second hit. The study suggests that failures of cell cycle control may be a common source of the large-scale chromosomal changes that dismantle multiple adaptation barriers during tumor evolution. Complex genome change, on this view, rarely arrives as a leisurely accumulation of small steps; it tends to arrive as a burst, born of a momentarily failing control system, and natural selection then decides what, if anything, survives.

Subject of Research: Loss of heterozygosity and aneuploidy driven by cell cycle control defects in diploid budding yeast

Article Title: Burst of aneuploidy: a cost of adaptation driven by breakdown of cell cycle control

Article References: Pirog, A., Tutaj, H., Tomala, K., & Korona, R. (2026). Burst of aneuploidy: a cost of adaptation driven by breakdown of cell cycle control. Molecular Systems Biology, 22(8), 1292-1311. https://doi.org/10.1038/s44320-026-00217-6

Image Credits: AI Generated

DOI: 10.1038/s44320-026-00217-6

Keywords: aneuploidy, loss of heterozygosity, cell cycle checkpoints, Saccharomyces cerevisiae, genome instability, mitosis, chromosome segregation, DNA repair, mutation rate, experimental evolution, cancer, genome rearrangements