Biology subjects

Eraslan, G.

Publications and source records attributed to Eraslan, G..

At least 19 recordsLinked to original sources

bioRxiv2026

GEMS (GWAS Effect Mechanisms via Sequence-to-Function): Unveiling cell type-specific mechanisms for GWAS variants

Genome-wide association studies (GWAS) have linked thousands of loci to common diseases, yet for non-coding variants, which constitute the vast majority of the signal, the cell types and states through which individual risk variants act remain largely unknown. Here, we develop GEMS (GWAS Effect Mechanisms via Sequence-to-Function), a framework that trains supervised classifiers on multimodal variant effect predictions (ATAC-seq and RNA-seq) from the sequence-to-function model Borzoi and applies SHAP-based attribution to associate individual GWAS variants with the cell types most relevant to their disease association. Applied to asthma, LDL cholesterol, multiple sclerosis, and type 2 diabetes, GEMS prioritizes known disease-relevant cell types concordant with stratified LD-score regression. At the variant level, GEMS associates a median of 58.7% of fine-mapped variants with at least one cell type via ATAC-seq and 65.9% via RNA-seq, extending cellular annotation beyond the chromatin accessibility peaks in the single-cell atlas used for model training. ATAC-seq variant-cell-type associations are significantly enriched in cis-regulatory elements from independent experimental studies not used in model training, and enrichment persists after excluding variants overlapping training atlas peaks. For multiple sclerosis, hierarchical biclustering of variant--cell-type associations partitions risk variants into immune and CNS modules driven by distinct transcription factor families, recapitulating the immune-mediated and neurodegenerative components of the disease and demonstrating that integrative analysis of sequence-to-function predictions can stratify phenotype-associated variants into mechanistically distinct, cell-type-specific regulatory programs.

genetics↗

bioRxiv2026

Deep learning design and in vivo validation of Müller glia-specific cis-regulatory elements

Effective recombinant adeno-associated virus gene therapies require promoters that are compact and cell-type-specific. Ideally, promoters should also exhibit functional conservation when tested in model organisms to ensure that preclinical findings translate reliably to human patients. Here, we introduce a deep learning framework for designing cis-regulatory elements (CREs) meeting these criteria, applied to retinal Muller glia (MG). Using single-cell chromatin accessibility data from human and mouse retinas, we trained species-specific models to predict cell-type accessibility, and designed compact CREs using two complementary strategies. In silico validation predicted that the designed CREs exhibit high MG-specific accessibility (on-target) in both species with minimal off-target accessibility across hundreds of human cell types and tissues. Mechanistic analysis revealed that the predicted MG accessibility is driven by the creation of LHX2 motifs. In vivo validation confirmed that the designed CREs successfully restrict reporter expression to MG in the murine retina. Our deep learning framework is highly generalizable and enables the rapid design of compact, on-target, and species-conserved CREs for precision gene therapy.

genomics↗

bioRxiv2026

gRely: Relyability for genome trained sequence-to-expression models

Sequence-to-function (S2F) models predict molecular phenotypes from DNA sequence and are increasingly applied to variant effect prediction (VEP), where the goal is to quantify how genetic variants alter gene expression. However, S2F model predictions are not uniformly reliable: accuracy varies substantially across variants, genes, and tissues, and current practice relies on crude magnitude thresholding to enrich for trustworthy predictions, which discards the majority of variants where S2F models could still provide signal. We developed gRely, a meta-modeling framework that estimates the probability that a given Borzoi VEP correctly predicts eQTL direction, using 1,121 features derived from the target variant, gene, and model outputs. On held-out tissues, gRely achieves a mean average precision of 0.885 (random baseline 0.744). Critically, within the low-magnitude regime where thresholding fails entirely, gRely identifies a high-confidence subset with 76% accuracy compared to a 58% baseline, recovering reliable predictions that magnitude filtering would discard. Interpretation via SHAP reveals that in this low-magnitude regime, gene expression level and cross-replicate signal concentration replace VEP magnitude as the primary discriminators of reliability. gRely is the first framework to provide per-prediction confidence scores for S2F model VEPs, and generalizes across architectures, producing consistent improvements on AlphaGenome predictions. By making reliability quantifiable, gRely enables principled filtering rather than blanket thresholding, and marks a step toward trustworthy deployment of S2F models in genomic research and clinical applications.

genomics↗

bioRxiv2026

Agentic systems are adept at solving well-scoped, verifiable problems in computational biology

We introduce CompBioBench, a benchmark of 100 diverse tasks for evaluating agentic systems in computational biology. Unlike mathematics and programming, which more readily admit systematic verification, biological data are inherently noisy and open to interpretation. To enable objective evaluation without reducing tasks to prescriptive checklists, we propose a new benchmark construction strategy based on synthetic/augmented data and metadata scrambling/scrubbing of real datasets to create challenging problems that have a single ground-truth answer and require multi-step reasoning, tool use, bespoke code, and interaction with real-world external resources. The benchmark spans genomics, transcriptomics, epigenomics, single-cell analysis, human genetics, and machine learning workflows. Questions are curated by domain experts to cover a broad range of skills with varying difficulty. We evaluate leading general-purpose agentic systems starting from a bare-minimum environment, requiring them to fetch data and tools as needed to solve each problem. We find strong end-to-end performance, with Codex CLI (GPT 5.4) reaching 83% accuracy, Gemini CLI (3.1 Pro) reaching 82%, Claude Code (Opus 4.6) reaching 81%, and Claude Code (Opus 4.7) reaching 78%. On the hardest questions, Claude Code (Opus 4.6) reaches 69%, Codex CLI (GPT 5.4) reaches 59%, and Gemini CLI (3.1 Pro) reaches 49%. CompBioBench provides a practical testbed for measuring the progress of agentic systems in computational biology and for guiding future benchmark design. Data and a public leaderboard are available at https://huggingface.co/collections/Genentech/compbiobench-v1.

bioinformatics↗

bioRxiv2025

Nona: A unifying multimodal masking framework for functional genomics

The non-coding genome encodes complex regulatory logic that orchestrates gene expression and cell identity. While machine learning models for functional genomics have advanced our understanding of the cis-regulatory code, sequence-to-function models, DNA language models, and generative models have evolved as separate paradigms despite probing the same underlying regulatory biology. We introduce Nona, a multimodal masked modeling framework that unifies these paradigms by learning jointly from DNA sequence and base-resolution functional genomics data. Beyond unifying existing modeling paradigms, Nona enables entirely new modeling objectives. We demonstrate its versatility through three applications: (1) a context-aware sequence-to-function model that improves local predictions by up to 13% by correcting systematic errors in sequence-to-function predictions; (2) a functional language model that integrates functional data into language modeling, learns relevant regulatory sequence motifs, and enables regulatory element design through masked discrete diffusion; (3) functional genotyping, which reveals an unrecognized privacy vulnerability in processed ATAC-seq data and re-identifies individuals from genetic databases with perfect accuracy. Together, these results establish masking as a universal interface for integrated modeling of functional genomics data, unifying disparate approaches while opening new directions for understanding and engineering the regulatory genome.

genomics↗

bioRxiv2025

Benchmarking DNA Sequence Models for Causal Regulatory Variant Prediction in Human Genetics

Machine learning holds immense promise in biology, particularly for the challenging task of identifying causal variants for Mendelian and complex traits. Two primary approaches have emerged for this task: supervised sequence-to-function models trained on functional genomics experimental data and self-supervised DNA language models that learn evolutionary constraints on sequences. However, the field currently lacks consistently curated datasets with accurate labels, especially for non-coding variants, that are necessary to comprehensively benchmark these models and advance the field. In this work, we present TraitGym, a curated dataset of regulatory genetic variants that are either known to be causal or are strong candidates across 113 Mendelian and 83 complex traits, along with carefully constructed control variants. We frame the causal variant prediction task as a binary classification problem and benchmark various models, including functional-genomics-supervised models, self-supervised models, models that combine machine learning predictions with curated annotation features, and ensembles of these. Our results provide insights into the capabilities and limitations of different approaches for predicting the functional consequences of non-coding genetic variants. We find that alignment-based models CADD and GPN-MSA compare favorably for Mendelian traits and complex disease traits, while functional-genomics-supervised models Enformer and Borzoi perform better for complex non-disease traits. Evo2 shows substantial performance gains with scale, but still lags somewhat behind alignment-based models, struggling particularly with enhancer variants. The benchmark, including a Google Colab notebook to evaluate a model in a few minutes, is available at https://huggingface.co/datasets/songlab/TraitGym.

genetics↗

bioRxiv2024

Decoding sequence determinants of gene expression in diverse cellular and disease states

Sequence-to-function models that predict gene expression from genomic DNA sequence have proven valuable for many biological tasks, including understanding cis-regulatory syntax and interpreting non-coding genetic variants. However, current state-of-the-art models have been trained largely on bulk expression profiles from healthy tissues or cell lines, and have not learned the properties of precise cell types and states that are captured in large-scale single-cell transcriptomic datasets. Thus, they lack the ability to perform these tasks at the resolution of specific cell types or states across diverse tissue and disease contexts. To address this gap, we present Decima, a model that predicts the cell type- and condition- specific expression of a gene from its surrounding DNA sequence. Decima is trained on single-cell or single-nucleus RNA sequencing data from over 22 million cells, and successfully predicts the cell type-specific expression of unseen genes based on their sequence alone. Here, we demonstrate Decimas ability to reveal the cis-regulatory mechanisms driving cell type-specific gene expression and its changes in disease, to predict non-coding variant effects at cell type resolution, and to design regulatory DNA elements with precisely tuned, context-specific functions.

genomics↗

bioRxiv2024

gReLU: A comprehensive framework for DNA sequence modeling and design

Deep learning models are increasingly being used to perform a variety of tasks on DNA sequences, such as predicting tissue- and cell type-specific sequence activity, deriving cis-regulatory rules, predicting non-coding variant effects, and designing synthetic regulatory sequences. However, these models require specialized knowledge to build, train and interpret correctly. In addition, the field is hampered by the lack of interoperability between models and software built by different groups. Here we present gReLU, a comprehensive software framework that enables users to easily execute advanced sequence modeling pipelines, including data preprocessing, model training, hyperparameter tuning, evaluation, interpretation, variant effect prediction, and design of novel regulatory elements. The software is accompanied by a model zoo containing state-of-the-art pre-trained models that can easily be downloaded, applied, and fine-tuned. This framework and resources will accelerate research across the field of DNA sequence modeling and enable the effective design of synthetic regulatory elements.

genomics↗

bioRxiv2024

regLM: Designing realistic regulatory DNA with autoregressive language models

Cis-regulatory elements (CREs), such as promoters and en-hancers, are DNA sequences that regulate the expression of genes. The activity of a CRE is influenced by the order, composition and spacing of sequence motifs that bind to proteins called transcription factors (TFs). Synthetic CREs with specific properties are needed for biomanufacturing as well as for many therapeutic applications including cell and gene therapy. Here, we present regLM, a framework to design synthetic CREs with desired properties, such as high, low or cell type-specific activity, using autoregressive language models in conjunction with supervised sequence-to-function models. We used our framework to design synthetic yeast promoters and cell type-specific human enhancers. We demonstrate that the synthetic CREs generated by our approach are not only predicted to have the desired functionality but also contain biological features similar to experimentally validated CREs. regLM thus facilitates the design of realistic regulatory DNA elements while providing insights into the cis-regulatory code.

genomics↗

bioRxiv2023

Polygraph: A Software Framework for the Systematic Assessment of Synthetic Regulatory DNA Elements

The design of regulatory elements is pivotal in gene and cell therapy, where DNA sequences are engineered to drive elevated and cell-type specific expression. However, the systematic assessment of synthetic DNA sequences without robust metrics and easy-to-use software remains challenging. Here, we introduce Polygraph, a Python framework that evaluates synthetic DNA elements, based on features like diversity, motif and k-mer composition, similarity to endogenous sequences, and screening with predictive and foundational models. Polygraph is the first instrument for assessing synthetic regulatory sequences, enabling faster progress in therapeutic interventions and improving our understanding of gene regulatory mechanisms.

bioinformatics↗

bioRxiv2023

Multi-modal skin atlas associates a multicellular immune-stromal community with altered cornification and T cell expansion in atopic dermatitis

In healthy skin, a cutaneous immune system maintains the balance between tolerance towards innocuous environmental antigens and immune responses against pathological agents. In atopic dermatitis (AD), barrier and immune dysfunction result in chronic tissue inflammation. Our understanding of the skin tissue ecosystem in AD remains incomplete with regard to the hallmarks of pathological barrier formation, and cellular state and clonal composition of disease-promoting cells. Here, we generated a multi-modal cell census of 310,691 cells spanning 86 cell subsets from whole skin tissue of 19 adult individuals, including non-lesional and lesional skin from 11 AD patients, and integrated it with 396,321 cells from four studies into a comprehensive human skin cell atlas in health and disease. Reconstruction of human keratinocyte differentiation from basal to cornified layers revealed a disrupted cornification trajectory in AD. This disrupted epithelial differentiation was associated with signals from a unique immune and stromal multicellular community comprised of MMP12+ dendritic cells (DCs), mature migratory DCs, cycling ILCs, NK cells, inflammatory CCL19+ IL4I1+ fibroblasts, and clonally expanded IL13+IL22+IL26+ T cells with overlapping type 2 and type 17 characteristics. Cell subsets within this immune and stromal multicellular community were connected by multiple inter-cellular positive feedback loops predicted to impact community assembly and maintenance. AD GWAS gene expression was enriched both in disrupted cornified keratinocytes and in cell subsets from the lesional immune and stromal multicellular community including IL13+IL22+IL26+ T cells and ILCs, suggesting that epithelial or immune dysfunction in the context of the observed cellular communication network can initiate and then converge towards AD. Our work highlights specific, disease-associated cell subsets and interactions as potential targets in progression and resolution of chronic inflammation.

molecular biology↗

bioRxiv2022

ChromDMM: A Dirichlet-Multinomial Mixture Model For Clustering Heterogeneous Epigenetic Data

MotivationResearch on epigenetic modifications and other chromatin features at genomic regulatory elements elucidates essential biological mechanisms including the regulation of gene expression. Despite the growing number of epigenetic datasets, new tools are still needed to discover novel distinctive patterns of heterogeneous epigenetic signals at regulatory elements. ResultsWe introduce ChromDMM, a product Dirichlet-multinomial mixture model for clustering genomic regions that are characterised by multiple chromatin features. ChromDMM extends the mixture model framework by profile shifting and flipping that can probabilistically account for inaccuracies in the position and strand-orientation of the genomic regions. Owing to hyper-parameter optimisation, ChromDMM can also regularise the smoothness of the epigenetic profiles across the consecutive genomic regions. With simulated data, we demonstrate that ChromDMM clusters, shifts, and strand-orients the profiles more accurately than previous methods. With ENCODE data, we show that the clustering of enhancer regions in the human genome reveals distinct patterns in several chromatin features. We further validate the enhancer clusters by their enrichment for transcriptional regulatory factor binding sites. AvailabilityThe software is available at https://github.com/MariaOsmala/ChromDMM

bioinformatics↗

bioRxiv2021

ECLIPSER: identifying causal cell types and genes for complex traits through single cell enrichment of e/sQTL-mapped genes in GWAS loci

SummaryECLIPSER was developed to identify pathogenic cell types and cell type-specific genes that may affect complex disease susceptibility and trait variation by integrating single cell data with known GWAS loci. ECLIPSER maps genes to GWAS loci for a given complex trait based on expression and splicing quantitative trait loci (e/sQTLs) and other functional data, and tests whether the mapped genes are enriched for cell type-specific expression in particular cell types using single-cell/nucleus RNA-seq data from one or more tissues of interest. A Bayesian Fishers exact test is used to compute fold-enrichment significance. We demonstrate the application of ECLIPSER on various skin diseases and traits using snRNA-seq of healthy human skin samples. Availability and ImplementationThe source code and documentation for ECLIPSER and a Jupyter notebook for generating output tables and figures are available at https://github.com/segrelabgenomics/ECLIPSER. The source code for GWASvar2gene that maps genes to GWAS loci based on e/sQTLs is available at https://github.com/segrelabgenomics/GWASvar2gene. The analysis presented here used data from GTEx (https://gtexportal.org/home/datasets) and Open Targets Genetics (https://genetics-docs.opentargets.org/data-access/graphql-api), but can also be applied to other GWAS variant lists and QTL studies. Data used to reproduce the results of the paper are available in Supplementary data.

bioinformatics↗

bioRxiv2021

Single-nucleus cross-tissue molecular reference maps to decipher disease gene function

Understanding the function of genes and their regulation in tissue homeostasis and disease requires knowing the cellular context in which genes are expressed in tissues across the body. Single cell genomics allows the generation of detailed cellular atlases in human tissues, but most efforts are focused on single tissue types. Here, we establish a framework for profiling multiple tissues across the human body at single-cell resolution using single nucleus RNA-Seq (snRNA-seq), and apply it to 8 diverse, archived, frozen tissue types (three donors per tissue). We apply four snRNA-seq methods to each of 25 samples from 16 donors, generating a cross-tissue atlas of 209,126 nuclei profiles, and benchmark them vs. scRNA-seq of comparable fresh tissues. We use a conditional variational autoencoder (cVAE) to integrate an atlas across tissues, donors, and laboratory methods. We highlight shared and tissue-specific features of tissue-resident immune cells, identifying tissue-restricted and non-restricted resident myeloid populations. These include a cross-tissue conserved dichotomy between LYVE1- and HLA class II-expressing macrophages, and the broad presence of LAM-like macrophages across healthy tissues that is also observed in disease. For rare, monogenic muscle diseases, we identify cell types that likely underlie the neuromuscular, metabolic, and immune components of these diseases, and biological processes involved in their pathology. For common complex diseases and traits analyzed by GWAS, we identify the cell types and gene modules that potentially underlie disease mechanisms. The experimental and analytical frameworks we describe will enable the generation of large-scale studies of how cellular and molecular processes vary across individuals and populations.

cell biology↗

bioRxiv2021

Single Cell, Single Nucleus and Spatial RNA Sequencing of the Human Liver Identifies Hepatic Stellate Cell and Cholangiocyte Heterogeneity

The critical functions of the human liver are coordinated through the interactions of hepatic parenchymal and non-parenchymal cells. Recent advances in single cell transcriptional approaches have enabled an examination of the human liver with unprecedented resolution. However, dissociation related cell perturbation can limit the ability to fully capture the human livers parenchymal cell fraction, which limits the ability to comprehensively profile this organ. Here, we report the transcriptional landscape of 73,295 cells from the human liver using matched single-cell RNA sequencing (scRNA-seq) and single-nucleus RNA sequencing (snRNA-seq). The addition of snRNA-seq enabled the characterization of interzonal hepatocytes at single-cell resolution, revealed the presence of rare subtypes of hepatic stellate cells previously only seen in disease, and detection of cholangiocyte progenitors that had only been observed during in vitro differentiation experiments. However, T and B lymphocytes and NK cells were only distinguishable using scRNA-seq, highlighting the importance of applying both technologies to obtain a complete map of tissue-resident cell-types. We validated the distinct spatial distribution of the hepatocyte, cholangiocyte and stellate cell populations by an independent spatial transcriptomics dataset and immunohistochemistry. Our study provides a systematic comparison of the transcriptomes captured by scRNA-seq and snRNA-seq and delivers a high-resolution map of the parenchymal cell populations in the healthy human liver.

cell biology↗

bioRxiv2020

Single-cell multimodal profiling of proteins and chromatin accessibility using PHAGE-ATAC

Multi-modal measurements of single cell profiles are a powerful tool for characterizing cell states and regulatory mechanisms. While current methods allow profiling of RNA along with either chromatin or protein levels, connecting chromatin state to protein levels remains a barrier. Here, we developed PHAGE-ATAC, a method that uses engineered camelid single-domain antibody ( nanobody)-displaying phages for simultaneous single-cell measurement of surface proteins, chromatin accessibility profiles, and mtDNA-based clonal tracing through a massively parallel droplet-based assay of single-cell transposase-accessible chromatin with sequencing (ATAC-seq). We demonstrate PHAGE-ATAC for multimodal analysis in primary human immune cells and for sample multiplexing. Finally, we construct a synthetic high-complexity phage library for selection of novel antigen-specific nanobodies that bind cells of particular molecular profiles, opening a new avenue for protein detection, cell characterization and screening with single-cell genomics.

genomics↗

bioRxiv2020

Integrated analyses of single-cell atlases reveal age, gender, and smoking status associations with cell type-specific expression of mediators of SARS-CoV-2 viral entry and highlights inflammatory programs in putative target cells

The COVID-19 pandemic, caused by the novel coronavirus SARS-CoV-2, creates an urgent need for identifying molecular mechanisms that mediate viral entry, propagation, and tissue pathology. Cell membrane bound angiotensin-converting enzyme 2 (ACE2) and associated proteases, transmembrane protease serine 2 (TMPRSS2) and Cathepsin L (CTSL), were previously identified as mediators of SARS-CoV2 cellular entry. Here, we assess the cell type-specific RNA expression of ACE2, TMPRSS2, and CTSL through an integrated analysis of 107 single-cell and single-nucleus RNA-Seq studies, including 22 lung and airways datasets (16 unpublished), and 85 datasets from other diverse organs. Joint expression of ACE2 and the accessory proteases identifies specific subsets of respiratory epithelial cells as putative targets of viral infection in the nasal passages, airways, and alveoli. Cells that co-express ACE2 and proteases are also identified in cells from other organs, some of which have been associated with COVID-19 transmission or pathology, including gut enterocytes, corneal epithelial cells, cardiomyocytes, heart pericytes, olfactory sustentacular cells, and renal epithelial cells. Performing the first meta-analyses of scRNA-seq studies, we analyzed 1,176,683 cells from 282 nasal, airway, and lung parenchyma samples from 164 donors spanning fetal, childhood, adult, and elderly age groups, associate increased levels of ACE2, TMPRSS2, and CTSL in specific cell types with increasing age, male gender, and smoking, all of which are epidemiologically linked to COVID-19 susceptibility and outcomes. Notably, there was a particularly low expression of ACE2 in the few young pediatric samples in the analysis. Further analysis reveals a gene expression program shared by ACE2+TMPRSS2+ cells in nasal, lung and gut tissues, including genes that may mediate viral entry, subtend key immune functions, and mediate epithelial-macrophage cross-talk. Amongst these are IL6, its receptor and co-receptor, IL1R, TNF response pathways, and complement genes. Cell type specificity in the lung and airways and smoking effects were conserved in mice. Our analyses suggest that differences in the cell type-specific expression of mediators of SARS-CoV-2 viral entry may be responsible for aspects of COVID-19 epidemiology and clinical course, and point to putative molecular pathways involved in disease susceptibility and pathogenesis.

bioinformatics↗

bioRxiv2019

High-density spatial transcriptomics arrays for in situ tissue profiling

Tissue function relies on the precise spatial organization of cells characterized by distinct molecular profiles. Single-cell RNA-Seq captures molecular profiles but not spatial organization. Conversely, spatial profiling assays either lack global transcriptome information or are not at the single-cell level. Here, we develop High-Density Spatial Transcriptomics (HDST), a method for RNA-seq at high spatial resolution. Spatially barcoded reverse transcription oligonucleotides are coupled to beads that are then randomly deposited in individual wells on a slide. The barcoded beads are decoded and coupled to a specific spatial address. We then capture and spatially in situ label RNA from the same histological tissue sections placed on the bead array slide. HDST recovers hundreds of thousands of transcript-coupled barcodes per experiment at 2 m resolution. We demonstrate HDST in the mouse brain, use it to resolve spatial expression patterns and cell types, and show how to combine it with histological stains to relate expression patterns to tissue architecture and anatomy. HDST opens the way to 2D spatial analysis of tissues at high resolution.

genomics↗