Biology subjects

Koo, P. K.

Publications and source records attributed to Koo, P. K..

18 recordsLinked to original sources

bioRxiv2026

Genetic background shapes AI-predicted variant effects

Predicting the consequences of genetic variants remains a major goal in biomedicine. Conventional approaches typically assess single-nucleotide variants in the context of a single reference genome, without accounting for genetic diversity that can modulate variant effects. Here we introduce the personalized variant effect predictor (pVEP) framework, which quantifies how genetic background across thousands of human genomes from globally diverse populations shapes computational predictions of clinical variant effects. Across deep learning models spanning protein structure, splicing, and noncoding regulation, pVEP reveals that many clinical variants exhibit heterogeneous predicted effects across haplotypes, with the same variant predicted to be pathogenic in some genetic backgrounds and benign in others. We find support for underlying molecular mechanisms, including shifts in predicted protein contacts and changes in splice-site recognition. Overall, personalized genomic context emerges as a systematically underappreciated variable in variant annotation and clinical interpretation, with particular implications for genetically diverse populations.

genomics↗

bioRxiv2024

Uncertainty-aware genomic deep learning with knowledge distillation

Deep neural networks (DNNs) have advanced predictive modeling for regulatory genomics, but challenges remain in ensuring the reliability of their predictions and understanding the key factors behind their decision making. Here we introduce DEGU (Distilling Ensembles for Genomic Uncertainty-aware models), a method that integrates ensemble learning and knowledge distillation to improve the robustness and explainability of DNN predictions. DEGU distills the predictions of an ensemble of DNNs into a single model, capturing both the average of the ensembles predictions and the variability across them, with the latter representing epistemic (or model-based) uncertainty. DEGU also includes an optional auxiliary task to estimate aleatoric, or data-based, uncertainty by modeling variability across experimental replicates. By applying DEGU across various functional genomic prediction tasks, we demonstrate that DEGU-trained models inherit the performance benefits of ensembles in a single model, with improved generalization to out-of-distribution sequences and more consistent explanations of cis-regulatory mechanisms through attribution analysis. Moreover, DEGU-trained models provide calibrated uncertainty estimates, with conformal prediction offering coverage guarantees under minimal assumptions. Overall, DEGU paves the way for robust and trustworthy applications of deep learning in genomics research.

genomics↗

bioRxiv2024

Evaluating the representational power of pre-trained DNA language models for regulatory genomics

The emergence of genomic language models (gLMs) offers an unsupervised approach to learning a wide diversity of cis-regulatory patterns in the non-coding genome without requiring labels of functional activity generated by wet-lab experiments. Previous evaluations have shown that pre-trained gLMs can be leveraged to improve predictive performance across a broad range of regulatory genomics tasks, albeit using relatively simple benchmark datasets and baseline models. Since the gLMs in these studies were tested upon fine-tuning their weights for each downstream task, determining whether gLM representations embody a foundational understanding of cis-regulatory biology remains an open question. Here we evaluate the representational power of pre-trained gLMs to predict and interpret cell-type-specific functional genomics data that span DNA and RNA regulation. Our findings suggest that probing the representations of pre-trained gLMs do not offer substantial advantages over conventional machine learning approaches that use one-hot encoded sequences. This work highlights a major gap with current gLMs, raising potential issues in conventional pre-training strategies for the non-coding genome.

bioinformatics↗

bioRxiv2023

Interpreting cis-regulatory mechanisms from genomic deep neural networks using surrogate models

Deep neural networks (DNNs) have greatly advanced the ability to predict genome function from sequence. Interpreting genomic DNNs in terms of biological mechanisms, however, remains difficult. Here we introduce SQUID, a genomic DNN interpretability framework based on surrogate modeling. SQUID approximates genomic DNNs in user-specified regions of sequence space using surrogate models, i.e., simpler models that are mechanistically interpretable. Importantly, SQUID removes the confounding effects that nonlinearities and heteroscedastic noise in functional genomics data can have on model interpretation. Benchmarking analysis on multiple genomic DNNs shows that SQUID, when compared to established interpretability methods, identifies motifs that are more consistent across genomic loci and yields improved single-nucleotide variant-effect predictions. SQUID also supports surrogate models that quantify epistatic interactions within and between cis-regulatory elements. SQUID thus advances the ability to mechanistically interpret genomic DNNs.

genomics↗

bioRxiv2023

Interpreting Cis-Regulatory Interactions from Large-Scale Deep Neural Networks for Genomics

The rise of large-scale, sequence-based deep neural networks (DNNs) for predicting gene expression has introduced challenges in their evaluation and interpretation. Current evaluations align DNN predictions with experimental perturbation assays, which provides insights into the generalization capabilities within the studied loci but offers a limited perspective of what drives their predictions. Moreover, existing model explainability tools focus mainly on motif analysis, which becomes complex when interpreting longer sequences. Here we introduce CREME, an in silico perturbation toolkit that interrogates large-scale DNNs to uncover rules of gene regulation that it learns. Using CREME, we investigate Enformer, a prominent DNN in gene expression prediction, revealing cis-regulatory elements (CREs) that directly enhance or silence target genes. We explore the intricate complexity of higher-order CRE interactions, the relationship between CRE distance from transcription start sites on gene expression, as well as the biochemical features of enhancers and silencers learned by Enformer. Moreover, we demonstrate the flexibility of CREME to efficiently uncover a higher-resolution view of functional sequence elements within CREs. This work demonstrates how CREME can be employed to translate the powerful predictions of large-scale DNNs to study open questions in gene regulation.

genomics↗

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↗

bioRxiv2021

End-to-end learning of multiple sequence alignments with differentiable Smith-Waterman

Multiple Sequence Alignments (MSAs) of homologous sequences contain information on structural and functional constraints and their evolutionary histories. Despite their importance for many downstream tasks, such as structure prediction, MSA generation is often treated as a separate pre-processing step, without any guidance from the application it will be used for. Here, we implement a smooth and differentiable version of the Smith-Waterman pairwise alignment algorithm that enables jointly learning an MSA and a downstream machine learning system in an end-to-end fashion. To demonstrate its utility, we introduce SMURF (Smooth Markov Unaligned Random Field), a new method that jointly learns an alignment and the parameters of a Markov Random Field for unsupervised contact prediction. We find that SMURF learns MSAs that mildly improve contact prediction on a diverse set of protein and RNA families. As a proof of concept, we demonstrate that by connecting our differentiable alignment module to AlphaFold and maximizing predicted confidence, we can learn MSAs that improve structure predictions over the initial MSAs. Interestingly, the alignments that improve AlphaFold predictions are self-inconsistent and can be viewed as adversarial. This work highlights the potential of differentiable dynamic programming to improve neural network pipelines that rely on an alignment and the potential dangers of relying on black-box methods for optimizing predictions of protein sequences.

bioinformatics↗

bioRxiv2021

Designing Interpretable Convolution-Based Hybrid Networks for Genomics

Hybrid networks that build upon convolutional layers with attention mechanisms have demon-strated improved performance relative to pure convolutional networks across many regulatory genome analysis tasks. Their inductive bias to learn long-range interactions provides an avenue to identify learned motif-motif interactions. For attention maps to be interpretable, the convolutional layer(s) must learn identifiable motifs. Here we systematically investigate the extent that architectural choices in convolution-based hybrid networks influence learned motif representations in first layer filters, as well as the reliability of their attribution maps generated by saliency analysis. We find that design principles previously identified in standard convolutional networks also generalize to hybrid networks. This work provides an avenue to narrow the spectrum of architectural choices when designing hybrid networks such that they are amenable to commonly used interpretability methods in genomics.

bioinformatics↗

bioRxiv2021

Exploiting marker genes for robust classification and characterization of single-cell chromatin accessibility

MotivationSingle-cell Assay for Transposase Accessible Chromatin using sequencing (scATAC-seq) is a valuable resource to learn cis-regulatory elements such as cell-type specific enhancers and transcription factor binding sites. However, cell-type identification of scATAC-seq data is known to be challenging due to the heterogeneity derived from different protocols and the high dropout rate. ResultsIn this study, we perform a systematic comparison of 7 scATAC-seq datasets of mouse brain to benchmark the efficacy of neuronal cell-type annotation from gene sets. We find that redundant marker genes give a dramatic improvement for a sparse scATAC-seq annotation across the data collected from different studies. Interestingly, simple aggregation of such marker genes achieves performance comparable or higher than that of machine-learning classifiers, suggesting its potential for downstream applications. Based on our results, we reannotated all scATAC-seq data for detailed cell types using robust marker genes. Their meta scATAC-seq profiles are publicly available at https://gillisweb.cshl.edu/Meta_scATAC. Furthermore, we trained a deep neural network to predict chromatin accessibility from only DNA sequence and identified key motifs enriched for each neuronal subtype. Those predicted profiles are visualized together in our database as a valuable resource to explore cell-type specific epigenetic regulation in a sequence-dependent and -independent manner. Contactjesse.gillis@utoronto.ca Supplementary informationSupplementary data are available at xxxxxx online. Key pointsO_LI7 scATAC-seq datasets of mouse brain are systematically compared to benchmark the efficacy of neuronal cell-type annotation from gene sets. C_LIO_LIRedundant marker genes give a dramatic improvement for a sparse scATAC-seq annotation beyond the heterogeneity of scATAC-seq data. C_LIO_LIWe reannotated all scATAC-seq data for detailed cell types using robust marker genes and their meta scATAC-seq profiles are publicly available at a new Meta scATAC-seq server. C_LIO_LIPredicted profiles from only DNA sequence using a deep neural network are visualized together to explore sequence-dependent and -independent epigenetic regulation. C_LI

bioinformatics↗

bioRxiv2020

Global Importance Analysis: A Method to QuantifyImportance of Genomic Features in Deep NeuralNetworks

Deep neural networks have demonstrated improved performance at predicting the sequence specificities of DNA- and RNA-binding proteins compared to previous methods that rely on k-mers and position weight matrices. To gain insights into why a DNN makes a given prediction, model interpretability methods, such as attribution methods, can be employed to identify motif-like representations along a given sequence. Because explanations are given on an individual sequence basis and can vary substantially across sequences, deducing generalizable trends across the dataset and quantifying their effect size remains a challenge. Here we introduce global importance analysis (GIA), a model interpretability method that quantifies the population-level effect size that putative patterns have on model predictions. GIA provides an avenue to quantitatively test hypotheses of putative patterns and their interactions with other patterns, as well as map out specific functions the network has learned. As a case study, we demonstrate the utility of GIA on the computational task of predicting RNA-protein interactions from sequence. We first introduce a convolutional network, we call ResidualBind, and benchmark its performance against previous methods on RNAcompete data. Using GIA, we then demonstrate that in addition to sequence motifs, ResidualBind learns a model that considers the number of motifs, their spacing, and sequence context, such as RNA secondary structure and GC-bias.

genomics↗

bioRxiv2020

Improving representations of genomic sequence motifs in convolutional networks with exponential activations

Deep convolutional neural networks (CNNs) trained on regulatory genomic sequences tend to build representations in a distributed manner, making it a challenge to extract learned features that are biologically meaningful, such as sequence motifs. Here we perform a comprehensive analysis on synthetic sequences to investigate the role that CNN activations have on model interpretability. We show that employing an exponential activation to first layer filters consistently leads to interpretable and robust representations of motifs compared to other commonly used activations. Strikingly, we demonstrate that CNNs with better test performance do not necessarily imply more interpretable representations with attribution methods. We find that CNNs with exponential activations significantly improve the efficacy of recovering biologically meaningful representations with attribution methods. We demonstrate these results generalise to real DNA sequences across several in vivo datasets. Together, this work demonstrates how a small modification to existing CNNs, i.e. setting exponential activations in the first layer, can significantly improve the robustness and interpretabilty of learned representations directly in convolutional filters and indirectly with attribution methods.

genomics↗

bioRxiv2020

Interpreting Deep Neural Networks Beyond Attribution Methods: Quantifying Global Importance of Genomic Features

Despite deep neural networks (DNNs) having found great success at improving performance on various prediction tasks in computational genomics, it remains difficult to understand why they make any given prediction. In genomics, the main approaches to interpret a high-performing DNN are to visualize learned representations via weight visualizations and attribution methods. While these methods can be informative, each has strong limitations. For instance, attribution methods only uncover the independent contribution of single nucleotide variants in a given sequence. Here we discuss and argue for global importance analysis which can quantify population-level importance of putative features and their interactions learned by a DNN. We highlight recent work that has benefited from this interpretability approach and then discuss connections between global importance analysis and causality.

bioinformatics↗

bioRxiv2019

Robust Neural Networks are More Interpretable for Genomics

Deep neural networks (DNNs) have been applied to a variety of regulatory genomics tasks. For interpretability, attribution methods are employed to provide importance scores for each nucleotide in a given sequence. However, even with state-of-the-art DNNs, there is no guarantee that these methods can recover interpretable, biological representations. Here we perform systematic experiments on synthetic genomic data to raise awareness of this issue. We find that deeper networks have better generalization performance, but attribution methods recover less interpretable representations. Then, we show training methods promoting robustness - including regularization, injecting random noise into the data, and adversarial training - significantly improve interpretability of DNNs, especially for smaller datasets.

genomics↗

bioRxiv2019

Improving Convolutional Network Interpretability with Exponential Activations

Deep convolutional networks trained on regulatory genomic sequences tend to learn distributed representations of sequence motifs across many first layer filters. This makes it challenging to decipher which features are biologically meaningful. Here we introduce the exponential activation that - when applied to first layer filters - leads to more interpretable representations of motifs, both visually and quantitatively, compared to rectified linear units. We demonstrate this on synthetic DNA sequences which have ground truth with various convolutional networks, and then show that this phenomenon holds on in vivo DNA sequences.

bioinformatics↗

bioRxiv2018

An Empirical Demonstration of Unsupervised Machine Learning in Species Delimitation

One major challenge to delimiting species with genetic data is successfully differentiating species divergences from population structure, with some current methods biased towards overestimating species numbers. Many fields of science are now utilizing machine learning (ML) approaches, and in systematics and evolutionary biology, supervised ML algorithms have recently been incorporated to infer species boundaries. However, these methods require the creation of training data with associated labels. Unsupervised ML, on the other hand, uses the inherent structure in data and hence does not require any user-specified training labels, thus providing a more objective approach to species delimitation. In the context of integrative taxonomy, we demonstrate the utility of three unsupervised ML approaches, specifically random forests, variational autoencoders, and t-distributed stochastic neighbor embedding, for species delimitation utilizing a short-range endemic harvestman taxon (Laniatores, Metanonychus). First, we combine mitochondrial data with examination of male genitalic morphology to identify a priori species hypotheses. Then we use single nucleotide polymorphism data derived from sequence capture of ultraconserved elements (UCEs) to test the efficacy of unsupervised ML algorithms in successfully identifying a priori species, comparing results to commonly used genetic approaches. Finally, we use two validation methods to assess a priori species hypotheses using UCE data. We find that unsupervised ML approaches successfully cluster samples according to species level divergences and not to high levels of population structure, while standard model-based validation methods over-split species, in some instances suggesting that all sampled individuals are distinct species. Moreover, unsupervised ML approaches offer the benefits of better data visualization in two-dimensional space and the ability to accommodate various data types. We argue that ML methods may be better suited for species delimitation relative to currently used model-based validation methods, and that species delimitation in a truly integrative framework provides more robust final species hypotheses relative to separating delimitation into distinct \"discovery\" and \"validation\" phases. Unsupervised ML is a powerful analytical approach that can be incorporated into many aspects of systematic biology, including species delimitation. Based on results of our empirical dataset, we make several taxonomic changes including description of a new species.

evolutionary biology↗

bioRxiv2018

Inferring Sequence-Structure Preferences of RNA-Binding Proteins with Convolutional Residual Networks

To infer the sequence and RNA structure specificities of RNA-binding proteins (RBPs) from experiments that enrich for bound sequences, we introduce a convolutional residual network which we call ResidualBind. ResidualBind significantly outperforms previous methods on experimental data from many RBP families. We interrogate ResidualBind to identify what features it has learned from high-affinity sequences with saliency analysis along with 1st-order and 2nd-order in silico mutagenesis. We show that in addition to sequence motifs, ResidualBind learns a model that includes the number of motifs, their spacing, and both positive and negative effects of RNA structure context. Strikingly, ResidualBind learns RNA structure context, including detailed base-pairing relationships, directly from sequence data, which we confirm on synthetic data. ResidualBind is a powerful, flexible, and interpretable model that can uncover cis-recognition preferences across a broad spectrum of RBPs.

genomics↗

bioRxiv2018

Representation Learning of Genomic Sequence Motifs with Convolutional Neural Networks

Although convolutional neural networks (CNNs) have been applied to a variety of computational genomics problems, there remains a large gap in our understanding of how they build representations of regulatory genomic sequences. Here we perform systematic experiments on synthetic sequences to reveal how CNN architecture, specifically convolutional filter size and max-pooling, influences the extent that sequence motif representations are learned by first layer filters. We find that CNNs designed to foster hierarchical representation learning of sequence motifs - assembling partial features into whole features in deeper layers - tend to learn distributed representations, i.e. partial motifs. On the other hand, CNNs that are designed to limit the ability to hierarchically build sequence motif representations in deeper layers tend to learn more interpretable localist representations, i.e. whole motifs. We then validate that this representation learning principle established from synthetic sequences generalizes to in vivo sequences.

genomics↗

bioRxiv2017

Inferring Functional Neural Connectivity With Deep Residual Convolutional Networks

Measuring synaptic connectivity in large neuronal populations remains a major goal of modern neuroscience. While this connectivity is traditionally revealed by anatomical methods such as electron microscopy, an efficient alternative is to computationally infer functional connectivity from recordings of neural activity. However, these statistical techniques still require further refinement before they can be reliably applied to real data. Here, we report significant improvements to a deep learning method for functional connectomics, as assayed on synthetic ChaLearn Connectomics data. The method, which integrates recent advances in convolutional neural network architecture and model-free partial correlation coefficients, outperforms published methods on competition data and can achieve over 90% precision at 1% recall on validation datasets. This suggests that future application of the model to in vivo whole-brain imaging data in larval zebrafish could reliably recover on the order of 106 synaptic connections with a 10% false discovery rate. The model also generalizes to networks with different underlying connection probabilities and should scale well when parallelized across multiple GPUs. The method offers real potential as a statistical complement to existing experiments and circuit hypotheses in neuroscience.

neuroscience↗