bioRxiv Science⌕ Search

Biology subjects

Siepel, A.

Publications and source records attributed to Siepel, A..

At least 19 recordsLinked to original sources
bioRxiv2026

Lineage-informed factor analysis reveals heritable programs of single-cell gene expression

Single-cell transcriptomics has transformed our ability to characterize cellular identity, but present-day gene expression profiles capture only a snapshot of a process that unfolds across cell division history. Recent single-cell lineage-tracing technologies make it possible to reconstruct cell division histories for thousands of cells, opening a window into how gene expression evolves. Yet observed gene expression is often redundant, with correlations among genes reflecting underlying latent biological programs and regulatory networks. To capture this structure, we introduce scPFA, a single-cell phylogenetic factor analysis framework that represents gene expression through a small set of latent factors that evolve along lineages under a phylogenetic prior, capturing correlated structure hidden in present-day observations alone. Simulations demonstrate accurate recovery of latent factors and covariance structure across conditions. Applied to developmental and cancer lineage-tracing datasets, the model uncovers biologically interpretable, lineage-associated expression programs. Together, these results demonstrate how lineage-informed factor analysis can reveal temporal biological structure hidden within high-dimensional single-cell data.

bioinformatics↗
bioRxiv2026

Lineage-aware stochastic modeling reveals gene-expression dynamics in development and disease

Gene expression changes along cell lineages, but most single-cell RNA-seq analyses treat cells as independent snapshots and ignore their phylogenetic relationships. Here we present LaVOUS, a lineage-aware probabilistic framework for modeling sparse single-cell gene-expression counts on reconstructed lineage trees. LaVOUS couples Brownian motion and Ornstein-Uhlenbeck models of latent transcriptional dynamics with negative-binomial observation models and scalable variational inference, enabling likelihood-based tests for gene-expression heritability, branch-specific expression shifts, and ancestral expression reconstruction. In simulations, LaVOUS improved detection of lineage-associated expression changes over Gaussian phylogenetic models and accurately reconstructed expression histories across expression levels. Applied to lineage-resolved single-cell datasets from metastatic lung cancer, class-switching B cells, and the developing brain, LaVOUS identified expression changes associated with metastatic progression, isotype switching, and neuronal differentiation. LaVOUS provides a general framework for studying single-cell expression dynamics across development and disease.

bioinformatics↗
bioRxiv2026

A comparative analysis of promoter-proximal pausing reveals kinetic and distributional dimensions of variation

Promoter-proximal pausing of RNA polymerase II is a key regulatory checkpoint in metazoan transcription. Despite extensive study of this process, quantitative methods for comparing pausing dynamics across biological contexts have been lacking. Here we introduce a model-based framework for rigorous comparative analysis of both pause-escape kinetics and pause-site distributions. We apply this framework to public datasets together with newly generated PRO-seq and Micro-C data from primate immune cells, enabling a broad comparison of promoter-proximal pausing across conditions, cell types, and species. Analyses of these datasets revealed striking differences across transcriptional perturbations and distinct patterns of variation in pause-escape kinetics and pause-site distributions across cell types and species, with only weak coupling between them. Integration with chromatin and sequence features showed that lower pause-escape rates are associated with stronger promoter-proximal nucleosome occupancy, whereas changes in pause-site dispersion are associated with sequence features such as GC skew. Together, these results reveal kinetic and distributional dimensions of pausing variation across biological contexts.

bioinformatics↗
bioRxiv2025

Variational Inference with Node Embeddings (VINE) for Scalable Bayesian Phylogenetics

Bayesian methods are now widely used in reconstructing both species and cell-lineage phylogenies, but they remain heavily reliant on computationally intensive Markov chain Monte Carlo sampling. Phylogenetic variational inference (VI) circumvents this dependency but so far has been limited in speed and scalability. Here we introduce Variational Inference with Node Embeddings (VO_SCPLOWINEC_SCPLOW), a computational method that combines an embedding of taxa in a high-dimensional space and a distance-based "decoder" with several algorithmic innovations to dramatically improve phylogenetic VI. VO_SCPLOWINEC_SCPLOW supports both standard DNA substitution models and CRISPR barcode-mutation models for inference of cell-lineage trees and tissue-migration histories. In extensive simulation experiments, we show that VO_SCPLOWINEC_SCPLOW can effectively approximate the results of the best available Bayesian methods with speeds orders of magnitude faster. We then apply VO_SCPLOWINEC_SCPLOW to [~]1,000 complete SARS-CoV-2 genomes and [~]900 lung-cancer cell barcodes, showing reductions in compute time from days to hours or minutes.

bioinformatics↗
bioRxiv2025

The persistence and loss of hard selective sweeps amid ancient human admixture

The extent to which human adaptations have persisted throughout history despite strong eroding demographic events such as admixture, genetic drift, and fluctuations in selection pressures remains unknown. Understanding which adaptations were resilient to such forces may shed light on traits that were important for humans across time. Yet, detecting selection from ancient DNA is challenging due to severe degradation of the data and/or signal. Here we detect selective sweeps using a domain-adaptive neural network (DANN) trained on simulated data and applied to more than 800 ancient and modern Eurasian genomes spanning the last 7000 years. We show that the DANN can account for simulation misspecification, or discrepancies between simulations and real aDNA, improving the ability to detect sweeps in real data compared to standard convolutional neural networks or standard statistics. Application of the DANN to data recovered 16 known sweeps at loci including LCT, HLA, KITLG, and OCA2/HERC2, and revealed 32 novel sweeps. All identified sweeps were classified as hard, consistent with historically low population sizes. While some sweeps were lost over time, 14 sweeps at loci involved in functions including neuronal, reproductive, pigmentation, and signaling traits persisted from the earliest to the most recent time periods. In most cases, the most frequent haplotype remained at high frequency across time. Together, these results indicate that hard sweeps predominated in ancient Eurasians and that several ancient selective events were resilient to strong admixture events. Significance statementThe extent to which human adaptations have persisted despite strong eroding forces such as admixture, drift, or fluctuations in selection pressures remains unknown. Understanding which loci are particularly resilient to such forces may shed light on the traits that were important for humans across time. Using a domain-adaptive neural network that accounts for simulations with misspecified demography relative to the data, we discover several sweeps at loci encoding neuronal, reproductive, pigmentation, and signaling traits persisted from the earliest time periods to the present, revealing the resilience of these sweeps to strong admixture events. Moreover, we find that hard sweeps, driven by single beneficial mutations, dominated throughout human history, consistent with the historically low human population sizes.

evolutionary biology↗
bioRxiv2025

Bayesian inference of tissue-migration histories in metastatic cancer from cell-lineage tracing data

Cell-lineage tracing now enables direct study of tissue migration in metastatic cancer, but current reconstruction algorithms are limited by a reliance on strong parsimony assumptions and pre-estimated cell-lineage phylogenies. Here, we introduce a probabilistic modeling and inference framework, called BEAM (Bayesian Evolutionary Analysis of Metastasis), that provides richer information about complex metastatic histories. Based on the flexible BEAST 2 platform for Bayesian phylogenetics, BEAM infers a full posterior distribution over cell-lineage phylogenies and tissue migration graphs, complete with timing information. We show using simulated data that BEAM reliably outperforms current methods for inference of tissue migration graphs, especially for more complex histories. We then apply BEAM to public data sets for lung and prostate cancer, finding support for distinct modes of migration across clones and reseeding of primary tumors. Overall, BEAM serves as a powerful framework for revealing the modes, timing, and directionality of tissue migration in metastatic cancer.

cancer biology↗
bioRxiv2025

GHIST 2024: The 1st Genomic History Inference Strategies Tournament

Evaluating population genetic inference methods is challenging due to the complexity of evolutionary histories, potential model misspecification, and unconscious biases in self-assessment. The Genomic History Inference Strategies Tournament (GHIST) is a community-driven competition designed to evaluate methods for inferring evolutionary history from population genomic data. The inaugural GHIST competition ran from July to November 2024 and featured four demographic history inference challenges of varying complexity: a bottleneck model, a split with isolation model, a secondary contact model with demographic complexity, and an archaic admixture model. Data were provided as error-free VCF files, and participants submitted numerical parameter estimates that were scored by relative root mean squared error. Approximately 60 participants competed, using diverse approaches. Results revealed the current dominance of methods based on site frequency spectra, while highlighting the advantages of flexible model-building approaches for complex demographic histories. We discuss insights regarding the competition and outline the next iteration, which is ongoing with expanded challenge diversity. By providing standardized benchmarks and highlighting areas for improvement, GHIST represents a substantial step toward more reliable inference of evolutionary history from genomic data.

evolutionary biology↗
bioRxiv2025

Extensive genome evolution distinguishes maize within a stable tribe of grasses

Over the last 20 million years, the Andropogoneae tribe of grasses has evolved to dominate 17% of global land area. Domestication of these grasses in the last 10,000 years has yielded our most productive crops, including maize, sugarcane, and sorghum. The majority of Andropogoneae species, including maize, show a history of polyploidy - a condition that, while offering the evolutionary advantage of multiple gene copies, poses challenges to basic cellular processes, gene expression, and epigenetic regulation. Genomic studies of polyploidy have been limited by sparse sampling of taxa in groups with multiple polyploidy events. Here, we present 33 genome assemblies from 27 species, including chromosome-scale assemblies of maize relatives Zea and Tripsacum. In maize, the after-effects of polyploidy have been widely studied, showing reduced chromosome number, biased fractionation of duplicate genes, and transposable element (TE) expansions. While we observe these patterns within the genus Zea, 12 other polyploidy events deviate significantly. Those tetraploids and hexaploids retain elevated chromosome number, maintain nearly complete complements of duplicate genes, and have only stochastic TE amplifications. These genomes reveal variable outcomes of polyploidy, challenging simple predictions and providing a foundation for understanding its evolutionary implications in an ecologically and economically important clade.

evolutionary biology↗
bioRxiv2024

Inferring polygenic negative selection underlying an individual trait as a distribution of fitness effects (DFEs) from GWAS summary statistics

There has been rising interest in exploiting data from genome-wide association studies (GWAS) to detect a genetic signature of natural selection acting on a given phenotype. However, current approaches are unable to directly estimate the distribution of fitness effects (DFE), an established property in population genetics that can elucidate genomic architecture pertaining to a particular focal trait. To this end, we introduce ASSESS, an inferential method that exploits the Poisson Random Field (PRF) to model selection coefficients from genome-wide allele count data, while jointly conditioning GWAS summary statistics on a latent distribution of phenotypic effect sizes. This probabilistic model is unified under the assumption of an explicit relationship between fitness and trait effect to yield a DFE. To gauge the performance of ASSESS, we enlisted various simulation experiments that covered a range of usage cases and model misspecifications, which revealed accurate recovery of the underlying selection signal. As a further proof-of-concept, ASSESS was applied to an array of publicly available human trait data, whereby we replicated previously published empirical findings from an alternative methodology. These demonstrations illustrate the potential of ASSESS to satisfy an increasing need for powerful yet convenient population genomic inference from GWAS summary statistics. Author SummaryThe growth of genome-wide association studies (GWAS) over the past decade has provided a wealth of resources for uncovering the genomic architecture underlying complex traits, including the footprint of selection. Currently, there are computational tools for inferring natural selection whereby GWAS results are leveraged to conduct a binary test for overall presence, estimate a correlated property, or summarize polygenic selection strength with a single statistic. However, a methodology that exploits GWAS data to estimate the distribution of fitness effects (DFE), which is the most direct measurement for the genetic impact of natural selection acting on a complex trait, does not currently exist. To this end, we constructed an approach to directly infer the DFE, wherein per-site selection coefficients specifically associated with a focal trait are aggregated across the genome. This implementation is designed to explicitly model an entire genome-wide set of summary statistics output from a GWAS rather than the individual-level input data, which offers computational efficiency and convenience as well as alleviates privacy concerns. We expect this to be a promising development given the further accumulation of GWAS results and investigators seeking more sophisticated analyses into the relationship between genetics and traits.

evolutionary biology↗
bioRxiv2023

Utility of AlphaMissense predictions in Asparagine Synthetase deficiency variant classification

AlphaMissense is a recently developed method that is designed to classify missense variants into pathogenic, benign, or ambiguous categories across the entire human proteome. Asparagine Synthetase Deficiency (ASNSD) is a developmental disorder associated with severe symptoms, including congenital microcephaly, seizures, and premature death. Diagnosing ASNSD relies on identifying mutations in the asparagine synthetase (ASNS) gene through DNA sequencing and determining whether these variants are pathogenic or benign. Pathogenic ASNS variants are predicted to disrupt the proteins structure and/or function, leading to asparagine depletion within cells and inhibition of cell growth. AlphaMissense offers a promising solution for the rapid classification of ASNS variants established by DNA sequencing and provides a community resource of pathogenicity scores and classifications for newly diagnosed ASNSD patients. Here, we assessed AlphaMissenses utility in ASNSD by benchmarking it against known critical residues in ASNS and evaluating its performance against a list of previously reported ASNSD-associated variants. We also present a pipeline to calculate AlphaMissense scores for any protein in the UniProt database. AlphaMissense accurately attributed a high average pathogenicity score to known critical residues within the two ASNS active sites and the connecting intramolecular tunnel. The program successfully categorized 78.9% of known ASNSD-associated missense variants as pathogenic. The remaining variants were primarily labeled as ambiguous, with a smaller proportion classified as benign. This study underscores the potential role of AlphaMissense in classifying ASNS variants in suspected cases of ASNSD, potentially providing clarity to patients and their families grappling with ongoing diagnostic uncertainty.

genetics↗
bioRxiv2023

Single-cell transcription mapping of murine and human mammary organoids responses to female hormones

During female adolescence and pregnancy, rising levels of hormones result in a cyclic source of signals that control the development of mammary tissue. While such alterations are well understood from a whole-gland perspective, the alterations that such hormones bring to organoid cultures derived from mammary glands have yet to be fully mapped. This is of special importance given that organoids are considered suitable systems to understand cross species breast development. Here we utilized single-cell transcriptional profiling to delineate responses of murine and human normal breast organoid systems to female hormones across evolutionary distinct species. Collectively, our study represents a molecular atlas of epithelial dynamics in response to estrogen and pregnancy hormones.

developmental biology↗
bioRxiv2023

Teosinte Pollen Drive guides maize domestication and evolution by RNAi

Meiotic drivers subvert Mendelian expectations by manipulating reproductive development to bias their own transmission. Chromosomal drive typically functions in asymmetric female meiosis, while gene drive is normally postmeiotic and typically found in males. Using single molecule and single-pollen genome sequencing, we describe Teosinte Pollen Drive, an instance of gene drive in hybrids between maize (Zea mays ssp. mays) and teosinte mexicana (Zea mays ssp. mexicana), that depends on RNA interference (RNAi). 22nt small RNAs from a non-coding RNA hairpin in mexicana depend on Dicer-Like 2 (Dcl2) and target Teosinte Drive Responder 1 (Tdr1), which encodes a lipase required for pollen viability. Dcl2, Tdr1, and the hairpin are in tight pseudolinkage on chromosome 5, but only when transmitted through the male. Introgression of mexicana into early cultivated maize is thought to have been critical to its geographical dispersal throughout the Americas, and a tightly linked inversion in mexicana spans a major domestication sweep in modern maize. A survey of maize landraces and sympatric populations of teosinte mexicana reveals correlated patterns of admixture among unlinked genes required for RNAi on at least 4 chromosomes that are also subject to gene drive in pollen from synthetic hybrids. Teosinte Pollen Drive likely played a major role in maize domestication and diversification, and offers an explanation for the widespread abundance of "self" small RNAs in the germlines of plants and animals.

plant biology↗
bioRxiv2023

Domain-adaptive neural networks improve supervised machine learning based on simulated population genetic data

Investigators have recently introduced powerful methods for population genetic inference that rely on supervised machine learning from simulated data. Despite their performance advantages, these methods can fail when the simulated training data does not adequately resemble data from the real world. Here, we show that this "simulation mis-specification" problem can be framed as a "domain adaptation" problem, where a model learned from one data distribution is applied to a dataset drawn from a different distribution. By applying an established domain-adaptation technique based on a gradient reversal layer (GRL), originally introduced for image classification, we show that the effects of simulation mis-specification can be substantially mitigated. We focus our analysis on two state-of-the-art deep-learning population genetic methods--SIA, which infers positive selection from features of the ancestral recombination graph (ARG), and ReLERNN, which infers recombination rates from genotype matrices. In the case of SIA, the domain adaptive framework also compensates for ARG inference error. Using the domain-adaptive SIA (dadaSIA) model, we estimate improved selection coefficients at selected loci in the 1000 Genomes CEU population. We anticipate that domain adaptation will prove to be widely applicable in the growing use of supervised machine learning in population genetics. Author SummaryPopulation genetic simulation is a powerful tool in the study of evolution. A number of supervised machine learning methods have been developed that take advantage of inexpensive simulations as training data. Despite their outstanding performance in benchmarks, these models can fail when the simulated training data deviate from the real data. In this work, we employed domain adaptation techniques to address this "simulation mis-specification" problem by training the machine learning model jointly with simulated and real data. We performed extensive benchmark experiments to demonstrate the improvement of the domain-adaptive models over standard machine learning models in the presence of different types of mis-specification. In addition, we applied dadaSIA, a domain-adaptive selection inference model, to improve the estimates of selection coefficients at selected loci in a European population. The domain adaptation framework proposed in our work is widely applicable to models relying on synthetic training data and therefore opens the door to many more applications in population genetics.

genetics↗
bioRxiv2022

Model-based characterization of the equilibrium dynamics of transcription initiation and promoter-proximal pausing in human cells

In metazoans, both transcription initiation and the escape of RNA polymerase (RNAP) from promoter-proximal pausing are key rate-limiting steps in gene expression. These processes play out at physically proximal sites on the DNA template and appear to influence one another through steric interactions, leading to a complex dynamic equilibrium in RNAP occupancy of the ~100 bp immediately downstream of the transcription start site. In this article, we examine the dynamics of these processes using a combination of statistical modeling, simulation, and analysis of real nascent RNA sequencing data. We develop a simple probabilistic model that jointly describes the kinetics of transcription initiation, pause-escape, and elongation, and the generation of nascent RNA sequencing read counts under steady-state conditions. We then extend this initial model to allow for variability across cells in promoter-proximal pause site locations and steric hindrance of transcription initiation from paused RNAPs. In an extensive series of simulations over a broad range of parameters, we show that this model enables accurate estimation of initiation and pause-escape rates even in the presence of collisions between RNAPs and variable elongation rates. Furthermore, we show by simulation and analysis of data for human cell lines that pause-escape is often more strongly rate-limiting than conventional "pausing indices" would suggest, that occupancy of the pause site is elevated at many genes, and that steric hindrance of initiation can lead to a pronounced reduction in apparent initiation rates. Our modeling framework is generally applicable for all types of nascent RNA sequencing data and can be applied to a variety of inference problems.

genomics↗
bioRxiv2022

RNA Pol II pausing facilitates phased pluripotency transitions by buffering transcription

Promoter-proximal RNA Pol II pausing is a critical step in transcriptional control. Pol II pausing has been studied predominantly in tissue culture systems. While Pol II pausing has been shown to be required for mammalian development, the phenotypic and mechanistic details of this requirement are unknown. Here, we find that loss of RNA Pol II pausing stalls pluripotent state transitions in the epiblast of the early mouse embryo. Using Nelfb-/- mice and a novel NELFB- degron mouse embryonic stem cells, we show that mouse ES cells (mESCs) representing the naive state of pluripotency successfully initiate a transition program, but fail to balance levels of induced and repressed genes and enhancers in the absence of NELF. Consistently, we find an increase in chromatin-associated NELF during pluripotency transitions. Overall, our work reveals the molecular and phenotypic roles of Pol II pausing in pluripotency and introduces Pol II pausing as a modulator of cell state transitions.

molecular biology↗
bioRxiv2021

Extreme purifying selection against point mutations in the human genome

Genome sequencing of tens of thousands of humans has enabled the measurement of large selective effects for mutations to protein-coding genes. Here we describe a new method, called ExtRaINSIGHT, for measuring similar selective effects in noncoding as well as in coding regions of the human genome. ExtRaINSIGHT estimates the prevalance of strong purifying selection, or "ultraselection" ({lambda}s), as the fractional depletion of rare single-nucleotide variants in target genomic sites relative to matched sites that are putatively free from selection, after controlling for local variation and neighbor-dependence in mutation rate. We show using simulations that{lambda} s is closely related to the average site-specific selection coefficient against heterozygous point mutations, as predicted at mutation-selection balance. Applying ExtRaINSIGHT to 71,702 whole genome sequences from gnomAD v3, we find strong evidence of ultraselection in evolutionarily ancient miRNAs and neuronal protein-coding genes, as well as at splice sites. By contrast, we find weak evidence in other noncoding RNAs and transcription factor binding sites, and only modest evidence in ultraconserved elements and human accelerated regions. We estimate that ~0.3-0.5% of the human genome is ultraselected, implying ~0.3-0.4 lethal or nearly lethal de novo mutations per potential human zygote. Overall, our study sheds new light on the genome-wide distribution of fitness effects for new point mutations by combining deep new sequencing data sets and classical theory from population genetics.

evolutionary biology↗
bioRxiv2021

SIA: Selection Inference Using the Ancestral Recombination Graph

Detecting signals of selection from genomic data is a central problem in population genetics. Coupling the rich information in the ancestral recombination graph (ARG) with a powerful and scalable deep learning framework, we developed a novel method to detect and quantify positive selection: Selection Inference using the Ancestral recombination graph (SIA). Built on a Long Short-Term Memory (LSTM) architecture, a particular type of a Recurrent Neural Network (RNN), SIA can be trained to explicitly infer a full range of selection coefficients, as well as the allele frequency trajectory and time of selection onset. We benchmarked SIA extensively on simulations under a European human demographic model, and found that it performs as well or better as some of the best available methods, including state-of-the-art machine-learning and ARG-based methods. In addition, we used SIA to estimate selection coefficients at several loci associated with human phenotypes of interest. SIA detected novel signals of selection particular to the European (CEU) population at the MC1R and ABCC11 loci. In addition, it recapitulated signals of selection at the LCT locus and several pigmentation-related genes. Finally, we reanalyzed polymorphism data of a collection of recently radiated southern capuchino seedeater taxa in the genus Sporophila to quantify the strength of selection and improved the power of our previous methods to detect partial soft sweeps. Overall, SIA uses deep learning to leverage the ARG and thereby provides new insight into how selective sweeps shape genomic diversity.

genetics↗