Biology subjects

Drusinsky, S.

Publications and source records attributed to Drusinsky, S..

4 recordsLinked to original sources

bioRxiv2026

Causal variant underestimation is a major overlooked driver of sequence-to-function model underperformance

Deep learning sequence-to-function (S2F) models represent tremendous promise for functionally fine-mapping causal variants associated with traits and disease. Yet it remains unclear precisely how effective they are at this task. Generally, S2F models perform well at classifying putatively causal expression quantitative trait loci (eQTL) SNVs, yet dramatically underperform linear baselines at ranking different individuals' gene expression values from their whole genome sequence, which directly calls into question their ability to fine-map a locus by properly weighting the effects of variants onto gene expression. Here, using the state-of-the-art S2F model AlphaGenome, we systematically compared predicted effect sizes for fine-mapped eQTLs to nearby putatively non-causal SNVs, finding that AlphaGenome pervasively underestimates the effects of most causal variants and fails to successfully fine-map eQTLs in most loci for this reason. We show that misdirected variant effect predictions and negative cross-individual correlations, widely cited as major challenges facing S2F expression modeling, are not egregious errors but a chance consequence of weak, noisy attributions when causal variants are underestimated. Our results suggest that a failure to detect local variant effects onto enhancer activity is a cause of underestimation. Additionally, we find that underestimated variants are enriched far from their target gene, suggesting inadequate enhancer-gene linking causes underestimation even when local effects are well-detected. Finally, our results clarify when S2F models are reliable for fine-mapping objectives: while they pervasively underestimate most causal variants, demonstrating limited fine-mapping utility for most loci, the top 0.1 most prominent variant effect predictions are strongly enriched for causal variants with directionally correct predictions that are consistent across model replicates, suggesting extremely strong predictions are broadly trustworthy. Our results demonstrate that causal variant underestimation is a core issue facing S2F expression predictors, with future improvements dependent on better local activity detection and enhancer-gene linking.

genetics↗

bioRxiv2025

GAME: Genomic API for Model Evaluation

The rapid expansion of genomics datasets and the application of machine learning has produced sequence-to-activity genomics models with ever-expanding capabilities. However, benchmarking these models on practical applications has been challenging because individual projects evaluate their models in ad hoc ways, and there is substantial heterogeneity of both model architectures and benchmarking tasks. To address this challenge, we have created GAME, a system for large-scale, community-led standardized model benchmarking on user-defined evaluation tasks. We borrow concepts from the Application Programming Interface (API) paradigm to allow for seamless communication between pre-trained models and benchmarking tasks, ensuring consistent evaluation protocols. Because all models and benchmarks are inherently compatible in this framework, the continual addition of new models and new benchmarks is easy. We also developed a Matcher module powered by a large language model (LLM) to automate ambiguous task alignment between benchmarks and models. Containerization of these modules enhances reproducibility and facilitates the deployment of models and benchmarks across computing platforms. By focusing on predicting underlying biochemical phenomena (e.g. gene expression, open chromatin, DNA binding), we ensure that tasks remain technology-independent. We provide examples of benchmarks and models implementing this framework, and anticipate that the community will contribute their own, leading to an ever-expanding and evolving set of models and evaluation tasks. This resource will accelerate genomics research by illuminating the best models for a given task, motivating novel functional genomic benchmarks, and providing a more nuanced understanding of model abilities.

bioinformatics↗

bioRxiv2024

Deep-learning prediction of gene expression from personal genomes

Models that predict RNA levels from DNA sequences show tremendous promise for decoding tissue-specific gene regulatory mechanisms1-5, revealing the genetic architecture of traits6-10, and interpreting noncoding genetic variation10,11. Existing methods take two different approaches: 1) associating expression with linear combinations of common genetic variants (training across individuals on single genes)12,13, or 2) learning genome-wide sequence-to-expression rules with neural networks (training across loci using a reference genome)11,14,15. Since limitations of both strategies have been highlighted recently16-20, we sought to combine the sequence context provided by deep learning with the information provided by cross-individual training. We utilized fine-tuning to develop Performer, a model with accuracy approaching the cis-heritability of most genes. Performer prioritizes genetic variants across the allele frequency spectrum that disrupt motifs, fall in annotated regulatory elements, and have functional evidence for modulating gene expression. While obstacles remain in personalized expression prediction, our findings establish deep learning as a viable strategy.

genetics↗

bioRxiv2020

Optimization and scaling of patient-derived brain organoids uncovers deep phenotypes of disease

Cerebral organoids provide unparalleled access to human brain development in vitro. However, variability induced by current culture methodologies precludes using organoids as robust disease models. To address this, we developed an automated Organoid Culture and Assay (ORCA) system to support longitudinal unbiased phenotyping of organoids at scale across multiple patient lines. We then characterized organoid variability using novel machine learning methods and found that the contribution of donor, clone, and batch is significant and remarkably consistent over gene expression, morphology, and cell-type composition. Next, we performed multi-factorial protocol optimization, producing a directed forebrain protocol compatible with 96-well culture that exhibits low variability while preserving tissue complexity. Finally, we used ORCA to study tuberous sclerosis, a disease with known genetics but poorly representative animal models. For the first time, we report highly reproducible early morphological and molecular signatures of disease in heterozygous TSC+/- forebrain organoids, demonstrating the benefit of a scaled organoid system for phenotype discovery in human disease models.

neuroscience↗