Biology subjects

Shrikumar, A.

Publications and source records attributed to Shrikumar, A..

15 recordsLinked to original sources

bioRxiv2024

Exploring Differences in Functional Connectivity in Australian Rules Football Players : A Resting-State fMRI Study on the Default Mode Network

The effects of non-concussive impacts in contact-sports such as in Australian rules football (ARF) are still largely unexplored. These impacts are often but not always lower in intensity, but occur more frequently than actual concussions. Since non-concussive impacts are often asymptomatic, their significance may be underestimated. Acute or subacute measurement of non-concussive injury is challenging as the pathological response and injury is poorly described. There is therefore a need for a greater understanding of the pathological consequences of exposure. Growing evidence indicates that resting-state functional connectivity (rs-fMRI) changes in the Default Mode Network (DMN) may be an important biomarker that is sensitive to characterize these impacts. In this work, we examined functional connectivity changes within the DMN of ARF players to evaluate its potential as an early biomarker for non-concussive impacts. Based on rs-fMRI, we compare the DMN of 47 sub-elite ARF players (mean age 21.5{+/-}2.7 years [SD], males 57%) and 42 age-matched healthy controls (mean age 23.2{+/-}2.3 years [SD], males 48%) using Independent Component Analysis (ICA) and Dual Regression. This approach permits an unbiased decomposition of brain activity into networks with principled handling of statistical error. An 83% increase in DMN connectivity (as measured by the Strictly Standardized Mean Difference on values derived from Dual Regression) was observed in ARF players in the left retrosplenial cingulate cortex compared to healthy controls (FDR-corrected p-value from dual regression = 0.03, 95% CI computed via bootstrapping was 58% to 116%). The AUC for distinguishing ARF players from controls was 0.80 (95% CI; [0.71, 0.89]), equating to a PPV of 78% and a NPV of 74%. These results are preliminary; future work could investigate robustness to different random initializations of ICA and validate the findings on an independent testing set, as well as investigate longitudinal changes in ARF players over the course of a playing season.

neuroscience↗

bioRxiv2024

Towards Principled Modeling Of Coronary Artery Calcium Scores With Zero-Inflated Regression

AO_SCPLOWBSTRACTC_SCPLOWThe Coronary Artery Calcium Score (CACS) is a widely-used measure of Coronary Artery Disease (CAD). The primary use of CACS is to identify subclinical CAD and estimate risk of future cardiovascular events such as acute myocardial infarction. The development of coronary atherosclerosis is well known to be accelerated in response to risk factors such as age, hypertension, hypercholeterolaemia and smoking. However, there is substantial variability in an individuals susceptibility or resistance to CAD against these risk factors. Quantifiying the deviation from "expected" CACS provides an novel opportunity to inform multi-omic and similar unbiased discovery of new markers and mechanisms of CAD trained against CT imaging. Standard linear regression struggles to model CACS due to the high prevalence of zeros (which reflect the absence of measurable coronary plaque). Prior works have variously handled this by discarding measurements with CACS of zero or by binning the CACS into broad categorical groups. Such approaches discard meaningful data, motivating the need for a more principled approach to handling the data distribution. In this work, we explored zero-inflated regression as a possible approach to modeling CACS using a cohort of patients from the BioHEART-CT study, and devised metrics to validate performance. We identified zero-inflated negative binomial regression, zero-inflated gamma regression and zero-inflated lognormal regression as promising approaches for handling the distributional properties of CACS, where the best method to use can vary depending on the dataset considered. A key contribution of our work is to demonstrate how these models can also estimate the percentile of the observed CACS relative to the distribution that would be expected after controlling for the inputs, thereby avoiding the need for data-hungry binning-based approaches.

bioinformatics↗

bioRxiv2023

De novo distillation of thermodynamic affinity from deep learning regulatory sequence models of in vivo protein-DNA binding

Transcription factors (TF) are proteins that bind DNA in a sequence-specific manner to regulate gene transcription. Despite their unique intrinsic sequence preferences, in vivo genomic occupancy profiles of TFs differ across cellular contexts. Hence, deciphering the sequence determinants of TF binding, both intrinsic and context-specific, is essential to understand gene regulation and the impact of regulatory, non-coding genetic variation. Biophysical models trained on in vitro TF binding assays can estimate intrinsic affinity landscapes and predict occupancy based on TF concentration and affinity. However, these models cannot adequately explain context-specific, in vivo binding profiles. Conversely, deep learning models, trained on in vivo TF binding assays, effectively predict and explain genomic occupancy profiles as a function of complex regulatory sequence syntax, albeit without a clear biophysical interpretation. To reconcile these complementary models of in vitro and in vivo TF binding, we developed Affinity Distillation (AD), a method that extracts thermodynamic affinities de-novo from deep learning models of TF chromatin immunoprecipitation (ChIP) experiments by marginalizing away the influence of genomic sequence context. Applied to neural networks modeling diverse classes of yeast and mammalian TFs, AD predicts energetic impacts of sequence variation within and surrounding motifs on TF binding as measured by diverse in vitro assays with superior dynamic range and accuracy compared to motif-based methods. Furthermore, AD can accurately discern affinities of TF paralogs. Our results highlight thermodynamic affinity as a key determinant of in vivo binding, suggest that deep learning models of in vivo binding implicitly learn high-resolution affinity landscapes, and show that these affinities can be successfully distilled using AD. This new biophysical interpretation of deep learning models enables high-throughput in silico experiments to explore the influence of sequence context and variation on both intrinsic affinity and in vivo occupancy.

genomics↗

bioRxiv2021

Towards More Realistic Simulated Datasets for Benchmarking Deep Learning Models in Regulatory Genomics

Deep neural networks and support vector machines have been shown to accurately predict genome-wide signals of regulatory activity from raw DNA sequences. These models are appealing in part because they can learn predictive DNA sequence features without prior assumptions. Several methods such as in-silico mutagenesis, GradCAM, DeepLIFT, Integrated Gradients and Gkm-Explain have been developed to reveal these learned features. However, the behavior of these methods on regulatory genomic data remains an area of active research. Although prior work has benchmarked these methods on simulated datasets with known ground-truth motifs, these simulations employed highly simplified regulatory logic that is not representative of the genome. In this work, we propose a novel pipeline for designing simulated data that comes closer to modeling the complexity of regulatory genomic DNA. We apply the pipeline to build simulated datasets based on publicly-available chromatin accessibility experiments and use these datasets to bench-mark different interpretation methods based on their ability to identify ground-truth motifs. We find that a GradCAM-based method, which was reported to perform well on a more simplified dataset, does not do well on this dataset (particularly when using an architecture with shorter convolutional kernels in the first layer), and we theoretically show that this is expected based on the nature of regulatory genomic data. We also show that Integrated Gradients sometimes performs worse than gradient-times-input, likely owing to its linear interpolation path. We additionally explore the impact of user-defined settings on the interpretation methods, such as the choice of "reference"/"baseline", and identify recommended settings for genomics. Our analysis suggests several promising directions for future research on these model interpretation methods. Code and links to data are available at https://github.com/kundajelab/interpret-benchmark.

bioinformatics↗

bioRxiv2021

Domain adaptive neural networks improvecross-species prediction of transcription factor binding

The intrinsic DNA sequence preferences and cell-type specific cooperative partners of transcription factors (TFs) are typically highly conserved. Hence, despite the rapid evolutionary turnover of individual TF binding sites, predictive sequence models of cell-type specific genomic occupancy of a TF in one species should generalize to closely matched cell types in a related species. To assess the viability of cross-species TF binding prediction, we train neural networks to discriminate ChIP-seq peak locations from genomic background and evaluate their performance within and across species. Cross-species predictive performance is consistently worse than within-species performance, which we show is caused in part by species-specific repeats. To account for this domain shift, we use an augmented network architecture to automatically discourage learning of training species-specific sequence features. This domain adaptation approach corrects for prediction errors on species-specific repeats and improves overall cross-species model performance. Our results demonstrate that cross-species TF binding prediction is feasible when models account for domain shifts driven by species-specific repeats.

bioinformatics↗

bioRxiv2020

Benchmarking Reverse-Complement Strategies for Deep Learning Models in Genomics

Predictive models mapping double-stranded DNA to signals of regulatory activity should, in principle, produce analogous (or "equivariant") predictions whether the forward strand or its reverse complement (RC) is supplied as input. Unfortunately, standard neural networks can produce highly divergent predictions across strands, even when the training set is augmented with RC sequences. Two strategies have emerged to enforce equivariance: conjoined/"siamese" architectures, and RC parameter sharing or RCPS. However, the connections between the two remain unclear, comparisons to strong baselines are lacking, and neither has been adapted to base-resolution signal profile prediction. Here we extend conjoined & RCPS models to base-resolution signal prediction, and introduce a strong baseline: a standard model (trained with RC data augmentation) that is made conjoined only after training, which we call "post-hoc" conjoined. Through benchmarks on diverse tasks, we find post-hoc conjoined consistently performs best or second-best, surpassed only occasionally by RCPS, and never underperforms conjoined-during-training. We propose an overfitting-based hypothesis for the latter finding, and study it empirically. Despite its theoretical appeal, RCPS shows mediocre performance on several tasks, even though (as we prove) it can represent any solution learned by conjoined models. Our results suggest users interested in RC equivariance should default to post-hoc conjoined as a reliable baseline before exploring RCPS. Finally, we present a unified description of conjoined & RCPS architectures, revealing a broader class of models that gradually interpolate between RCPS and conjoined while maintaining equivariance.

bioinformatics↗

bioRxiv2020

fastISM: Performant in-silico saturation mutagenesis for convolutional neural networks

Deep learning models such as convolutional neural networks are able to accurately map biological sequences to associated functional readouts and properties by learning predictive de novo representations. In-silico saturation mutagenesis (ISM) is a popular feature attribution technique for inferring contributions of all characters in an input sequence to the models predicted output. The main drawback of ISM is its runtime, as it involves multiple forward propagations of all possible mutations of each character in the input sequence through the trained model to predict the effects on the output. We present fastISM, an algorithm that speeds up ISM by a factor of over 10x for commonly used convolutional neural network architectures. fastISM is based on the observations that the majority of computation in ISM is spent in convolutional layers, and a single mutation only disrupts a limited region of intermediate layers, rendering most computation redundant. fastISM reduces the gap between backpropagation-based feature attribution methods and ISM. It far surpasses the runtime of backpropagation-based methods on multi-output architectures, making it feasible to run ISM on a large number of sequences. An easy-to-use Keras/TensorFlow 2 implementation of fastISM is available at https://github.com/kundajelab/fastISM, and a hands-on tutorial at https://colab.research.google.com/github/kundajelab/fastISM/blob/master/notebooks/colab/DeepSEA.ipynb.

bioinformatics↗

bioRxiv2020

Fourier-transform-based attribution priors improve the interpretability and stability of deep learning models for genomics

Deep learning models can accurately map genomic DNA sequences to associated functional molecular readouts such as protein-DNA binding data. Base-resolution importance (i.e. "attribution") scores inferred from these models can highlight predictive sequence motifs and syntax. Unfortunately, these models are prone to overfitting and are sensitive to random initializations, often resulting in noisy and irreproducible attributions that obfuscate underlying motifs. To address these shortcomings, we propose a novel attribution prior, where the Fourier transform of input-level attribution scores are computed at training-time, and high-frequency components of the Fourier spectrum are penalized. We evaluate different model architectures with and without attribution priors trained on genome-wide binary or continuous molecular profiles. We show that our attribution prior dramatically improves models stability, interpretability, and performance on held-out data, especially when training data is severely limited. Our attribution prior also allows models to identify biologically meaningful sequence motifs more sensitively and precisely within individual regulatory elements. The prior is agnostic to the model architecture or predicted experimental assay, yet provides similar gains across all experiments. This work represents an important advancement in improving the reliability of deep learning models for deciphering the regulatory code of the genome.

bioinformatics↗

bioRxiv2019

Deep learning at base-resolution reveals motif syntax of the cis-regulatory code

The arrangement of transcription factor (TF) binding motifs (syntax) is an important part of the cis-regulatory code, yet remains elusive. We introduce a deep learning model, BPNet, that uses DNA sequence to predict base-resolution ChIP-nexus binding profiles of pluripotency TFs. We develop interpretation tools to learn predictive motif representations and identify soft syntax rules for cooperative TF binding interactions. Strikingly, Nanog preferentially binds with helical periodicity, and TFs often cooperate in a directional manner, which we validate using CRISPR-induced point mutations. Our model represents a powerful general approach to uncover the motifs and syntax of cis-regulatory sequences in genomics data. HighlightsO_LIThe neural network BPNet accurately predicts TF binding data at base-resolution. C_LIO_LIModel interpretation discovers TF motifs and TF interactions dependent on soft syntax. C_LIO_LIMotifs for Nanog and partners are preferentially spaced at [~]10.5 bp periodicity. C_LIO_LIDirectional cooperativity is validated: Sox2 enhances Nanog binding, but not vice versa. C_LI

genomics↗

bioRxiv2018

Gkmexplain: Fast and Accurate Interpretation of Nonlinear Gapped k-mer Support Vector Machines Using Integrated Gradients

Support Vector Machines with gapped k-mer kernels (gkm-SVMs) have been used to learn predictive models of regulatory DNA sequence. However, interpreting predictive sequence patterns learned by gkm-SVMs can be challenging. Existing interpretation methods such as deltaSVM, in-silico mutagenesis (ISM), or SHAP either do not scale well or make limiting assumptions about the model that can produce misleading results when the gkm kernel is combined with nonlinear kernels. Here, we propose gkmexplain: a novel approach inspired by the method of Integrated Gradients for interpreting gkm-SVM models. Using simulated regulatory DNA sequences, we show that gkmexplain identifies predictive patterns with high accuracy while avoiding pitfalls of deltaSVM and ISM and being orders of magnitude more computationally efficient than SHAP. We use a novel motif discovery method called TF-MoDISco to recover consolidated TF motifs from gkm-SVM models of in vivo TF binding by aggregating predictive patterns identified by gkmexplain. Finally, we find that mutation impact scores derived through gkmexplain using gkm-SVM models of chromatin accessibility in lymphoblastoid cell-lines consistently outperform deltaSVM and ISM at identifying regulatory genetic variants (dsQTLs). Code and example notebooks replicating the workflow are available at https://github.com/kundajelab/gkmexplain. Explanatory videos available at http://bit.ly/gkmexplainvids.

bioinformatics↗

bioRxiv2018

Deciphering regulatory DNA sequences and noncoding genetic variants using neural network models of massively parallel reporter assays

The relationship between noncoding DNA sequence and gene expression is not well-understood. Massively parallel reporter assays (MPRAs), which quantify the regulatory activity of large libraries of DNA sequences in parallel, are a powerful approach to characterize this relationship. We present MPRA-DragoNN, a convolutional neural network (CNN)-based framework to predict and interpret the regulatory activity of DNA sequences as measured by MPRAs. While our method is generally applicable to a variety of MPRA designs, here we trained our model on the Sharpr-MPRA dataset that measures the activity of ~500,000 constructs tiling 15,720 regulatory regions in human K562 and HepG2 cell lines. MPRA-DragoNN predictions were moderately correlated (Spearman{rho} = 0.28) with measured activity and were within range of replicate concordance of the assay. State-of-the-art model interpretation methods revealed high-resolution predictive regulatory sequence features that overlapped transcription factor (TF) binding motifs. We used the model to investigate the cell type and chromatin state preferences of predictive TF motifs. We explored the ability of our model to predict the allelic effects of regulatory variants in an independent MPRA experiment and fine map putative functional SNPs in loci associated with lipid traits. Our results suggest that interpretable deep learning models trained on MPRA data have the potential to reveal meaningful patterns in regulatory DNA sequences and prioritize regulatory genetic variants, especially as larger, higher-quality datasets are produced.

genomics↗

bioRxiv2018

Kipoi: accelerating the community exchange and reuse of predictive models for genomics

Advanced machine learning models applied to large-scale genomics datasets hold the promise to be major drivers for genome science. Once trained, such models can serve as a tool to probe the relationships between data modalities, including the effect of genetic variants on phenotype. However, lack of standardization and limited accessibility of trained models have hampered their impact in practice. To address this, we present Kipoi, a collaborative initiative to define standards and to foster reuse of trained models in genomics. Already, the Kipoi repository contains over 2,000 trained models that cover canonical prediction tasks in transcriptional and post-transcriptional gene regulation. The Kipoi model standard grants automated software installation and provides unified interfaces to apply and interpret models. We illustrate Kipoi through canonical use cases, including model benchmarking, transfer learning, variant effect prediction, and building new models from existing ones. By providing a unified framework to archive, share, access, use, and build on models developed by the community, Kipoi will foster the dissemination and use of machine learning models in genomics.

bioinformatics↗

bioRxiv2017

Separable Fully Connected Layers Improve Deep Learning Models For Genomics

Convolutional neural networks are rapidly gaining popularity in regulatory genomics. Typically, these networks have a stack of convolutional and pooling layers, followed by one or more fully connected layers. In genomics, the same positional patterns are often present across multiple convolutional channels. Therefore, in current state-of-the-art networks, there exists significant redundancy in the representations learned by standard fully connected layers. We present a new separable fully connected layer that learns a weights tensor that is the outer product of positional weights and cross-channel weights, thereby allowing the same positional patterns to be applied across multiple convolutional channels. Decomposing positional and cross-channel weights further enables us to readily impose biologically-inspired constraints on positional weights, such as symmetry. We also propose a novel regularizer and constraint that act on curvature in the positional weights. Using experiments on simulated and in vivo datasets, we show that networks that incorporate our separable fully connected layer outperform conventional models with analogous architectures and the same number of parameters. Additionally, our networks are more robust to hyperparameter tuning, have more informative gradients, and produce importance scores that are more consistent with known biology than conventional deep neural networks.\n\nAvailabilityImplementation: https://github.com/kundajelab/keras/tree/keras_1\n\nA gist illustrating model setup is at: goo.gl/gYooaa

bioinformatics↗

bioRxiv2017

Opportunities And Obstacles For Deep Learning In Biology And Medicine

Deep learning, which describes a class of machine learning algorithms, has recently showed impressive results across a variety of domains. Biology and medicine are data rich, but the data are complex and often ill-understood. Problems of this nature may be particularly well-suited to deep learning techniques. We examine applications of deep learning to a variety of biomedical problems--patient classification, fundamental biological processes, and treatment of patients--and discuss whether deep learning will transform these tasks or if the biomedical sphere poses unique challenges. We find that deep learning has yet to revolutionize or definitively resolve any of these problems, but promising advances have been made on the prior state of the art. Even when improvement over a previous baseline has been modest, we have seen signs that deep learning methods may speed or aid human investigation. More work is needed to address concerns related to interpretability and how to best model each problem. Furthermore, the limited amount of labeled data for training presents problems in some domains, as do legal and privacy constraints on work with sensitive health records. Nonetheless, we foresee deep learning powering changes at both bench and bedside with the potential to transform several areas of biology and medicine.

bioinformatics↗

bioRxiv2017

Reverse-complement parameter sharing improves deep learning models for genomics

Deep learning approaches that have produced breakthrough predictive models in computer vision, speech recognition and machine translation are now being successfully applied to problems in regulatory genomics. However, deep learning architectures used thus far in genomics are often directly ported from computer vision and natural language processing applications with few, if any, domain-specific modifications. In double-stranded DNA, the same pattern may appear identically on one strand and its reverse complement due to complementary base pairing. Here, we show that conventional deep learning models that do not explicitly model this property can produce substantially different predictions on forward and reverse-complement versions of the same DNA sequence. We present four new convolutional neural network layers that leverage the reverse-complement property of genomic DNA sequence by sharing parameters between forward and reverse-complement representations in the model. These layers guarantee that forward and reverse-complement sequences produce identical predictions within numerical precision. Using experiments on simulated and in vivo transcription factor binding data, we show that our proposed architectures lead to improved performance, faster learning and cleaner internal representations compared to conventional architectures trained on the same data.\n\nAvailabilityOur implementation is available at\n\nContactavanti@stanford.edu, pgreens@stanford.edu, akundaje@stanford.edu

bioinformatics↗