Biology subjects

Toneyan, S.

Publications and source records attributed to Toneyan, S..

3 recordsLinked to original sources

bioRxiv2022

Evolution-inspired augmentations improve deep learning for regulatory genomics

Deep neural networks (DNNs) hold promise for functional genomics prediction, but their generalization capability may be limited by the amount of available data. To address this, we propose EvoAug, a suite of evolution-inspired augmentations that enhance the training of genomic DNNs by increasing genetic variation. However, random transformation of DNA sequences can potentially alter their function in unknown ways. Thus, we employ a fine-tuning procedure using the original non-transformed data to preserve functional integrity. Our results demonstrate that EvoAug substantially improves the generalization and interpretability of established DNNs across prominent regulatory genomics prediction tasks, offering a robust solution for genomic DNNs.

bioinformatics↗

bioRxiv2022

Evaluating deep learning for predicting epigenomic profiles

Deep learning has been successful at predicting epigenomic profiles from DNA sequences. Most approaches frame this task as a binary classification relying on peak callers to define functional activity. Recently, quantitative models have emerged to directly predict the experimental coverage values as a regression. As new models continue to emerge with different architectures and training configurations, a major bottleneck is forming due to the lack of ability to fairly assess the novelty of proposed models and their utility for downstream biological discovery. Here we introduce a unified evaluation framework and use it to compare various binary and quantitative models trained to predict chromatin accessibility data. We highlight various modeling choices that affect generalization performance, including a downstream application of predicting variant effects. In addition, we introduce a robustness metric that can be used to enhance model selection and improve variant effect predictions. Our empirical study largely supports that quantitative modeling of epigenomic profiles leads to better generalizability and interpretability.

bioinformatics↗

bioRxiv2021

Deconvolution of Expression for Nascent RNA Sequencing Data (DENR) Highlights Pre-RNA Isoform Diversity in Human Cells

Quantification of mature-RNA isoform abundance from RNA-seq data has been extensively studied, but much less attention has been devoted to quantifying the abundance of distinct precursor RNAs based on nascent RNA sequencing data. Here we address this problem with a new computational method called Deconvolution of Expression for Nascent RNA sequencing data (DENR). DENR models the nascent RNA read counts at each locus as a mixture of user-provided isoforms. The performance of the baseline algorithm is enhanced by the use of machine-learning predictions of transcription start sites (TSSs) and an adjustment for the typical "shape profile" of read counts along a transcription unit. We show using simulated data that DENR clearly outperforms simple read-count-based methods for estimating the abundances of both whole genes and isoforms. By applying DENR to previously published PRO-seq data from K562 and CD4+ T cells, we find that transcription of multiple isoforms per gene is widespread, and the dominant isoform frequently makes use of an internal TSS. We also identify > 200 genes whose dominant isoforms make use of different TSSs in these two cell types. Finally, we apply DENR and StringTie to newly generated PRO-seq and RNA-seq data, respectively, for human CD4+ T cells and CD14+ monocytes, and show that entropy at the pre-RNA level makes a disproportionate contribution to overall isoform diversity, especially across cell types. Altogether, DENR is the first computational tool to enable abundance quantification of pre-RNA isoforms based on nascent RNA sequencing data, and it reveals high levels of pre-RNA isoform diversity in human cells.

genomics↗