The Kipoi repository accelerates community exchange and reuse of predictive models for genomics

Small data methods in omics: the power of one

Over the last decade, biology has begun utilizing 'big data' approaches, resulting in large, comprehensive atlases in modalities ranging from transcriptomics to neural connectomics. However, these approaches must be complemented and integrated with 'small data' approaches to efficiently utilize data from individual labs. Integration of smaller datasets with major reference atlases is critical to provide context to individual experiments, and approaches toward integration of large and small data have been a major focus in many fields in recent years. Here we discuss progress in integration of small data with consortium-sized atlases across multiple modalities, and its potential applications. We then examine promising future directions for utilizing the power of small data to maximize the information garnered from small-scale experiments. We envision that, in the near future, international consortia comprising many laboratories will work together to collaboratively build reference atlases and foundation models using small data methods.

Deep Learning Sequence Models for Transcriptional Regulation

Deciphering the regulatory code of gene expression and interpreting the transcriptional effects of genome variation are critical challenges in human genetics. Modern experimental technologies have resulted in an abundance of data, enabling the development of sequence-based deep learning models that link patterns embedded in DNA to the biochemical and regulatory properties contributing to transcriptional regulation, including modeling epigenetic marks, 3D genome organization, and gene expression, with tissue and cell-type specificity. Such methods can predict the functional consequences of any noncoding variant in the human genome, even rare or never-before-observed variants, and systematically characterize their consequences beyond what is tractable from experiments or quantitative genetics studies alone. Recently, the development and application of interpretability approaches have led to the identification of key sequence patterns contributing to the predicted tasks, providing insights into the underlying biological mechanisms learned and revealing opportunities for improvement in future models.

Koina: Democratizing machine learning for proteomics research

Recent developments in machine-learning (ML) and deep-learning (DL) have immense potential for applications in proteomics, such as generating spectral libraries, improving peptide identification, and optimizing targeted acquisition modes. Although new ML/DL models for various applications and peptide properties are frequently published, the rate at which these models are adopted by the community is slow, which is mostly due to technical challenges. We believe that, for the community to make better use of state-of-the-art models, more attention should be spent on making models easy to use and accessible by the community. To facilitate this, we developed Koina, an open-source containerized, decentralized and online-accessible high-performance prediction service that enables ML/DL model usage in any pipeline. Using the widely used FragPipe computational platform as example, we show how Koina can be easily integrated with existing proteomics software tools and how these integrations improve data analysis.

Guidelines for releasing a variant effect predictor

Computational methods for assessing the likely impacts of mutations, known as variant effect predictors (VEPs), are widely used in the assessment and interpretation of human genetic variation, as well as in other applications like protein engineering. Many different VEPs have been released to date, and there is tremendous variability in their underlying algorithms and outputs, and in the ways in which the methodologies and predictions are shared. This leads to considerable challenges for end users in knowing which VEPs to use and how to use them. Here, to address these issues, we provide guidelines and recommendations for the release of novel VEPs. Emphasising open-source availability, transparent methodologies, clear variant effect score interpretations, standardised scales, accessible predictions, and rigorous training data disclosure, we aim to improve the usability and interpretability of VEPs, and promote their integration into analysis and evaluation pipelines. We also provide a large, categorised list of currently available VEPs, aiming to facilitate the discovery and encourage the usage of novel methods within the scientific community.

Cell-type-directed design of synthetic enhancers

Transcriptional enhancers act as docking stations for combinations of transcription factors and thereby regulate spatiotemporal activation of their target genes1. It has been a long-standing goal in the field to decode the regulatory logic of an enhancer and to understand the details of how spatiotemporal gene expression is encoded in an enhancer sequence. Here we show that deep learning models2-6, can be used to efficiently design synthetic, cell-type-specific enhancers, starting from random sequences, and that this optimization process allows detailed tracing of enhancer features at single-nucleotide resolution. We evaluate the function of fully synthetic enhancers to specifically target Kenyon cells or glial cells in the fruit fly brain using transgenic animals. We further exploit enhancer design to create 'dual-code' enhancers that target two cell types and minimal enhancers smaller than 50 base pairs that are fully functional. By examining the state space searches towards local optima, we characterize enhancer codes through the strength, combination and arrangement of transcription factor activator and transcription factor repressor motifs. Finally, we apply the same strategies to successfully design human enhancers, which adhere to enhancer rules similar to those of Drosophila enhancers. Enhancer design guided by deep learning leads to better understanding of how enhancers work and shows that their code can be exploited to manipulate cell states.

Predictive analyses of regulatory sequences with EUGENe

Deep learning has become a popular tool to study cis-regulatory function. Yet efforts to design software for deep-learning analyses in regulatory genomics that are findable, accessible, interoperable and reusable (FAIR) have fallen short of fully meeting these criteria. Here we present elucidating the utility of genomic elements with neural nets (EUGENe), a FAIR toolkit for the analysis of genomic sequences with deep learning. EUGENe consists of a set of modules and subpackages for executing the key functionality of a genomics deep learning workflow: (1) extracting, transforming and loading sequence data from many common file formats; (2) instantiating, initializing and training diverse model architectures; and (3) evaluating and interpreting model behavior. We designed EUGENe as a simple, flexible and extensible interface for streamlining and customizing end-to-end deep-learning sequence analyses, and illustrate these principles through application of the toolkit to three predictive modeling tasks. We hope that EUGENe represents a springboard towards a collaborative ecosystem for deep-learning applications in genomics research.

O’Brien T, Perucca P, Ge Z, Anderson A. Applications for Deep Learning in Epilepsy Genetic Research

Epilepsy is a group of brain disorders characterised by an enduring predisposition to generate unprovoked seizures. Fuelled by advances in sequencing technologies and computational approaches, more than 900 genes have now been implicated in epilepsy. The development and optimisation of tools and methods for analysing the vast quantity of genomic data is a rapidly evolving area of research. Deep learning (DL) is a subset of machine learning (ML) that brings opportunity for novel investigative strategies that can be harnessed to gain new insights into the genomic risk of people with epilepsy. DL is being harnessed to address limitations in accuracy of long-read sequencing technologies, which improve on short-read methods. Tools that predict the functional consequence of genetic variation can represent breaking ground in addressing critical knowledge gaps, while methods that integrate independent but complimentary data enhance the predictive power of genetic data. We provide an overview of these DL tools and discuss how they may be applied to the analysis of genetic data for epilepsy research.

Transfer Learning Allows Accurate RBP Target Site Prediction with Limited Sample Sizes

RNA-binding proteins are vital regulators in numerous biological processes. Their disfunction can result in diverse diseases, such as cancer or neurodegenerative disorders, making the prediction of their binding sites of high importance. Deep learning (DL) has brought about a revolution in various biological domains, including the field of protein-RNA interactions. Nonetheless, several challenges persist, such as the limited availability of experimentally validated binding sites to train well-performing DL models for the majority of proteins. Here, we present a novel training approach based on transfer learning (TL) to address the issue of limited data. Employing a sophisticated and interpretable architecture, we compare the performance of our method trained using two distinct approaches: training from scratch (SCR) and utilizing TL. Additionally, we benchmark our results against the current state-of-the-art methods. Furthermore, we tackle the challenges associated with selecting appropriate input features and determining optimal interval sizes. Our results show that TL enhances model performance, particularly in datasets with minimal training data, where satisfactory results can be achieved with just a few hundred RNA binding sites. Moreover, we demonstrate that integrating both sequence and evolutionary conservation information leads to superior performance. Additionally, we showcase how incorporating an attention layer into the model facilitates the interpretation of predictions within a biologically relevant context.

Short tandem repeats bind transcription factors to tune eukaryotic gene expression

Short tandem repeats (STRs) are enriched in eukaryotic cis-regulatory elements and alter gene expression, yet how they regulate transcription remains unknown. We found that STRs modulate transcription factor (TF)-DNA affinities and apparent on-rates by about 70-fold by directly binding TF DNA-binding domains, with energetic impacts exceeding many consensus motif mutations. STRs maximize the number of weakly preferred microstates near target sites, thereby increasing TF density, with impacts well predicted by statistical mechanics. Confirming that STRs also affect TF binding in cells, neural networks trained only on in vivo occupancies predicted effects identical to those observed in vitro. Approximately 90% of TFs preferentially bound STRs that need not resemble known motifs, providing a cis-regulatory mechanism to target TFs to genomic sites.

A curated census of pathogenic and likely pathogenic UTR variants and evaluation of deep learning models for variant effect prediction

Introduction: Variants in 5' and 3' untranslated regions (UTR) contribute to rare disease. While predictive algorithms to assist in classifying pathogenicity can potentially be highly valuable, the utility of these tools is often unclear, as it depends on carefully selected training and validation conditions. To address this, we developed a high confidence set of pathogenic (P) and likely pathogenic (LP) variants and assessed deep learning (DL) models for predicting their molecular effects. Methods: 3' and 5' UTR variants documented as P or LP (P/LP) were obtained from ClinVar and refined by reviewing the annotated variant effect and reassessing evidence of pathogenicity following published guidelines. Prediction scores from sequence-based DL models were compared between three groups: P/LP variants acting though the mechanism for which the model was designed (model-matched), those operating through other mechanisms (model-mismatched), and putative benign variants. PhyloP was used to compare conservation scores between P/LP and putative benign variants. Results: 295 3' and 188 5' UTR variants were obtained from ClinVar, of which 26 3' and 68 5' UTR variants were classified as P/LP. Predictions by DL models achieved statistically significant differences when comparing modelmatched P/LP variants to both putative benign variants and modelmismatched P/LP variants, as well as when comparing all P/LP variants to putative benign variants. PhyloP conservation scores were significantly higher among P/LP compared to putative benign variants for both the 3' and 5' UTR. Discussion: In conclusion, we present a high-confidence set of P/LP 3' and 5' UTR variants spanning a range of mechanisms and supported by detailed pathogenicity and molecular mechanism evidence curation. Predictions from DL models further substantiate these classifications. These datasets will support further development and validation of DL algorithms designed to predict the functional impact of variants that may be implicated in rare disease.

BioAutoMATED: An end-to-end automated machine learning tool for explanation and design of biological sequences

The design choices underlying machine-learning (ML) models present important barriers to entry for many biologists who aim to incorporate ML in their research. Automated machine-learning (AutoML) algorithms can address many challenges that come with applying ML to the life sciences. However, these algorithms are rarely used in systems and synthetic biology studies because they typically do not explicitly handle biological sequences (e.g., nucleotide, amino acid, or glycan sequences) and cannot be easily compared with other AutoML algorithms. Here, we present BioAutoMATED, an AutoML platform for biological sequence analysis that integrates multiple AutoML methods into a unified framework. Users are automatically provided with relevant techniques for analyzing, interpreting, and designing biological sequences. BioAutoMATED predicts gene regulation, peptide-drug interactions, and glycan annotation, and designs optimized synthetic biology components, revealing salient sequence characteristics. By automating sequence modeling, BioAutoMATED allows life scientists to incorporate ML more readily into their work.

Variant calling and benchmarking in an era of complete human genome sequences

Genetic variant calling from DNA sequencing has enabled understanding of germline variation in hundreds of thousands of humans. Sequencing technologies and variant-calling methods have advanced rapidly, routinely providing reliable variant calls in most of the human genome. We describe how advances in long reads, deep learning, de novo assembly and pangenomes have expanded access to variant calls in increasingly challenging, repetitive genomic regions, including medically relevant regions, and how new benchmark sets and benchmarking methods illuminate their strengths and limitations. Finally, we explore the possible future of more complete characterization of human genome variation in light of the recent completion of a telomere-to-telomere human genome reference assembly and human pangenomes, and we consider the innovations needed to benchmark their newly accessible repetitive regions and complex variants.

Current sequence-based models capture gene expression determinants in promoters but mostly ignore distal enhancers

BACKGROUND

The largest sequence-based models of transcription control to date are obtained by predicting genome-wide gene regulatory assays across the human genome. This setting is fundamentally correlative, as those models are exposed during training solely to the sequence variation between human genes that arose through evolution, questioning the extent to which those models capture genuine causal signals.

RESULTS

Here we confront predictions of state-of-the-art models of transcription regulation against data from two large-scale observational studies and five deep perturbation assays. The most advanced of these sequence-based models, Enformer, by and large, captures causal determinants of human promoters. However, models fail to capture the causal effects of enhancers on expression, notably in medium to long distances and particularly for highly expressed promoters. More generally, the predicted impact of distal elements on gene expression predictions is small and the ability to correctly integrate long-range information is significantly more limited than the receptive fields of the models suggest. This is likely caused by the escalating class imbalance between actual and candidate regulatory elements as distance increases.

CONCLUSIONS

Our results suggest that sequence-based models have advanced to the point that in silico study of promoter regions and promoter variants can provide meaningful insights and we provide practical guidance on how to use them. Moreover, we foresee that it will require significantly more and particularly new kinds of data to train models accurately accounting for distal elements.

The Network Zoo: a multilingual package for the inference and analysis of gene regulatory networks

Inference and analysis of gene regulatory networks (GRNs) require software that integrates multi-omic data from various sources. The Network Zoo (netZoo; netzoo.github.io) is a collection of open-source methods to infer GRNs, conduct differential network analyses, estimate community structure, and explore the transitions between biological states. The netZoo builds on our ongoing development of network methods, harmonizing the implementations in various computing languages and between methods to allow better integration of these tools into analytical pipelines. We demonstrate the utility using multi-omic data from the Cancer Cell Line Encyclopedia. We will continue to expand the netZoo to incorporate additional methods.

Harnessing the potential of machine learning and artificial intelligence for dementia research

Progress in dementia research has been limited, with substantial gaps in our knowledge of targets for prevention, mechanisms for disease progression, and disease-modifying treatments. The growing availability of multimodal data sets opens possibilities for the application of machine learning and artificial intelligence (AI) to help answer key questions in the field. We provide an overview of the state of the science, highlighting current challenges and opportunities for utilisation of AI approaches to move the field forward in the areas of genetics, experimental medicine, drug discovery and trials optimisation, imaging, and prevention. Machine learning methods can enhance results of genetic studies, help determine biological effects and facilitate the identification of drug targets based on genetic and transcriptomic information. The use of unsupervised learning for understanding disease mechanisms for drug discovery is promising, while analysis of multimodal data sets to characterise and quantify disease severity and subtype are also beginning to contribute to optimisation of clinical trial recruitment. Data-driven experimental medicine is needed to analyse data across modalities and develop novel algorithms to translate insights from animal models to human disease biology. AI methods in neuroimaging outperform traditional approaches for diagnostic classification, and although challenges around validation and translation remain, there is optimism for their meaningful integration to clinical practice in the near future. AI-based models can also clarify our understanding of the causality and commonality of dementia risk factors, informing and improving risk prediction models along with the development of preventative interventions. The complexity and heterogeneity of dementia requires an alternative approach beyond traditional design and analytical approaches. Although not yet widely used in dementia research, machine learning and AI have the potential to unlock current challenges and advance precision dementia medicine.

Deep learning in regulatory genomics: from identification to design

Genomics and deep learning are a natural match since both are data-driven fields. Regulatory genomics refers to functional noncoding DNA regulating gene expression. In recent years, deep learning applications on regulatory genomics have achieved remarkable advances so-much-so that it has revolutionized the rules of the game of the computational methods in this field. Here, we review two emerging trends: (i) the modeling of very long input sequence (up to 200 kb), which requires self-matched modularization of model architecture; (ii) on the balance of model predictability and model interpretability because the latter is more able to meet biological demands. Finally, we discuss how to employ these two routes to design synthetic regulatory DNA, as a promising strategy for optimizing crop agronomic properties.

Deep learning models will shape the future of stem cell research

Our ability to understand and control stem cell biology is being augmented by developments on two fronts, our ability to collect more data describing cell state and our capability to comprehend these data using deep learning models. Here we consider the impact deep learning will have in the future of stem cell research. We explore the importance of generating data suitable for these methods, the requirement for close collaboration between experimental and computational researchers, and the challenges we face to do this fairly and effectively. Achieving this will ensure that the resulting deep learning models are biologically meaningful and computationally tractable.

Computational prediction of human deep intronic variation

BACKGROUND

The adoption of whole-genome sequencing in genetic screens has facilitated the detection of genetic variation in the intronic regions of genes, far from annotated splice sites. However, selecting an appropriate computational tool to discriminate functionally relevant genetic variants from those with no effect is challenging, particularly for deep intronic regions where independent benchmarks are scarce.

RESULTS

In this study, we have provided an overview of the computational methods available and the extent to which they can be used to analyze deep intronic variation. We leveraged diverse datasets to extensively evaluate tool performance across different intronic regions, distinguishing between variants that are expected to disrupt splicing through different molecular mechanisms. Notably, we compared the performance of SpliceAI, a widely used sequence-based deep learning model, with that of more recent methods that extend its original implementation. We observed considerable differences in tool performance depending on the region considered, with variants generating cryptic splice sites being better predicted than those that potentially affect splicing regulatory elements. Finally, we devised a novel quantitative assessment of tool interpretability and found that tools providing mechanistic explanations of their predictions are often correct with respect to the ground - information, but the use of these tools results in decreased predictive power when compared to black box methods.

CONCLUSIONS

Our findings translate into practical recommendations for tool usage and provide a reference framework for applying prediction tools in deep intronic regions, enabling more informed decision-making by practitioners.

Deep learning approaches for noncoding variant prioritization in neurodegenerative diseases

Determining how noncoding genetic variants contribute to neurodegenerative dementias is fundamental to understanding disease pathogenesis, improving patient prognostication, and developing new clinical treatments. Next generation sequencing technologies have produced vast amounts of genomic data on cell type-specific transcription factor binding, gene expression, and three-dimensional chromatin interactions, with the promise of providing key insights into the biological mechanisms underlying disease. However, this data is highly complex, making it challenging for researchers to interpret, assimilate, and dissect. To this end, deep learning has emerged as a powerful tool for genome analysis that can capture the intricate patterns and dependencies within these large datasets. In this review, we organize and discuss the many unique model architectures, development philosophies, and interpretation methods that have emerged in the last few years with a focus on using deep learning to predict the impact of genetic variants on disease pathogenesis. We highlight both broadly-applicable genomic deep learning methods that can be fine-tuned to disease-specific contexts as well as existing neurodegenerative disease research, with an emphasis on Alzheimer's-specific literature. We conclude with an overview of the future of the field at the intersection of neurodegeneration, genomics, and deep learning.

Performance evaluation of differential splicing analysis methods and splicing analytics platform construction

A proportion of previously defined benign variants or variants of uncertain significance in humans, which are challenging to identify, may induce an abnormal splicing process. An increasing number of methods have been developed to predict splicing variants, but their performance has not been completely evaluated using independent benchmarks. Here, we manually sourced ∼50 000 positive/negative splicing variants from > 8000 studies and selected the independent splicing variants to evaluate the performance of prediction methods. These methods showed different performances in recognizing splicing variants in donor and acceptor regions, reminiscent of different weight coefficient applications to predict novel splicing variants. Of these methods, 66.67% exhibited higher specificities than sensitivities, suggesting that more moderate cut-off values are necessary to distinguish splicing variants. Moreover, the high correlation and consistent prediction ratio validated the feasibility of integration of the splicing prediction method in identifying splicing variants. We developed a splicing analytics platform called SPCards, which curates splicing variants from publications and predicts splicing scores of variants in genomes. SPCards also offers variant-level and gene-level annotation information, including allele frequency, non-synonymous prediction and comprehensive functional information. SPCards is suitable for high-throughput genetic identification of splicing variants, particularly those located in non-canonical splicing regions.

A review of deep learning applications in human genomics using next-generation sequencing data

Genomics is advancing towards data-driven science. Through the advent of high-throughput data generating technologies in human genomics, we are overwhelmed with the heap of genomic data. To extract knowledge and pattern out of this genomic data, artificial intelligence especially deep learning methods has been instrumental. In the current review, we address development and application of deep learning methods/models in different subarea of human genomics. We assessed over- and under-charted area of genomics by deep learning techniques. Deep learning algorithms underlying the genomic tools have been discussed briefly in later part of this review. Finally, we discussed briefly about the late application of deep learning tools in genomic. Conclusively, this review is timely for biotechnology or genomic scientists in order to guide them why, when and how to use deep learning methods to analyse human genomic data.

DeepSTARR predicts enhancer activity from DNA sequence and enables the de novo design of synthetic enhancers

Enhancer sequences control gene expression and comprise binding sites (motifs) for different transcription factors (TFs). Despite extensive genetic and computational studies, the relationship between DNA sequence and regulatory activity is poorly understood, and de novo enhancer design has been challenging. Here, we built a deep-learning model, DeepSTARR, to quantitatively predict the activities of thousands of developmental and housekeeping enhancers directly from DNA sequence in Drosophila melanogaster S2 cells. The model learned relevant TF motifs and higher-order syntax rules, including functionally nonequivalent instances of the same TF motif that are determined by motif-flanking sequence and intermotif distances. We validated these rules experimentally and demonstrated that they can be generalized to humans by testing more than 40,000 wildtype and mutant Drosophila and human enhancers. Finally, we designed and functionally validated synthetic enhancers with desired activities de novo.

Ad hoc learning of peptide fragmentation from mass spectra enables an interpretable detection of phosphorylated and cross-linked peptides

Mass spectrometry-based proteomics provides a holistic snapshot of the entire protein set of living cells on a molecular level. Currently, only a few deep learning approaches exist that involve peptide fragmentation spectra, which represent partial sequence information of proteins. Commonly, these approaches lack the ability to characterize less studied or even unknown patterns in spectra because of their use of explicit domain knowledge. Here, to elevate unrestricted learning from spectra, we introduce ‘ad hoc learning of fragmentation’ (AHLF), a deep learning model that is end-to-end trained on 19.2 million spectra from several phosphoproteomic datasets. AHLF is interpretable, and we show that peak-level feature importance values and pairwise interactions between peaks are in line with corresponding peptide fragments. We demonstrate our approach by detecting post-translational modifications, specifically protein phosphorylation based on only the fragmentation spectrum without a database search. AHLF increases the area under the receiver operating characteristic curve (AUC) by an average of 9.4% on recent phosphoproteomic data compared with the current state of the art on this task. Furthermore, use of AHLF in rescoring search results increases the number of phosphopeptide identifications by a margin of up to 15.1% at a constant false discovery rate. To show the broad applicability of AHLF, we use transfer learning to also detect cross-linked peptides, as used in protein structure analysis, with an AUC of up to 94%.

Supervised promoter recognition: a benchmark framework

Motivation

Deep learning has become a prevalent method in identifying genomic regulatory sequences such as promoters. In a number of recent papers, the performance of deep learning models has continually been reported as an improvement over alternatives for sequence-based promoter recognition. However, the performance improvements in these models do not account for the different datasets that models are evaluated on. The lack of a consensus dataset and procedure for benchmarking purposes has made the comparison of each model’s true performance difficult to assess.

Results

We present a framework called Supervised Promoter Recognition Framework (‘SUPR REF’) capable of streamlining the complete process of training, validating, testing, and comparing promoter recognition models in a systematic manner. SUPR REF includes the creation of biologically relevant benchmark datasets to be used in the evaluation process of deep learning promoter recognition models. We showcase this framework by comparing the models’ performances on alternative datasets, and properly evaluate previously published models on new benchmark datasets. Our results show that the reliability of deep learning ab initio promoter recognition models on eukaryotic genomic sequences is still not at a sufficient level, as overall performance is still low. These results originate from a subset of promoters, the well-known RNA Polymerase II core promoters. Furthermore, given the observational nature of these data, cross-validation results from small promoter datasets need to be interpreted with caution.

Computational and experimental methods for classifying variants of unknown clinical significance

The increase in sequencing capacity, reduction in costs, and national and international coordinated efforts have led to the widespread introduction of next-generation sequencing (NGS) technologies in patient care. More generally, human genetics and genomic medicine are gaining importance for more and more patients. Some communities are already discussing the prospect of sequencing each individual's genome at time of birth. Together with digital health records, this shall enable individualized treatments and preventive measures, so-called precision medicine. A central step in this process is the identification of disease causal mutations or variant combinations that make us more susceptible for diseases. Although various technological advances have improved the identification of genetic alterations, the interpretation and ranking of the identified variants remains a major challenge. Based on our knowledge of molecular processes or previously identified disease variants, we can identify potentially functional genetic variants and, using different lines of evidence, we are sometimes able to demonstrate their pathogenicity directly. However, the vast majority of variants are classified as variants of uncertain clinical significance (VUSs) with not enough experimental evidence to determine their pathogenicity. In these cases, computational methods may be used to improve the prioritization and an increasing toolbox of experimental methods is emerging that can be used to assay the molecular effects of VUSs. Here, we discuss how computational and experimental methods can be used to create catalogs of variant effects for a variety of molecular and cellular phenotypes. We discuss the prospects of integrating large-scale functional data with machine learning and clinical knowledge for the development of accurate pathogenicity predictions for clinical applications.

The evolution, evolvability and engineering of gene regulatory DNA

Mutations in non-coding regulatory DNA sequences can alter gene expression, organismal phenotype and fitness^1-3. Constructing complete fitness landscapes, in which DNA sequences are mapped to fitness, is a long-standing goal in biology, but has remained elusive because it is challenging to generalize reliably to vast sequence spaces^4-6. Here we build sequence-to-expression models that capture fitness landscapes and use them to decipher principles of regulatory evolution. Using millions of randomly sampled promoter DNA sequences and their measured expression levels in the yeast Saccharomyces cerevisiae, we learn deep neural network models that generalize with excellent prediction performance, and enable sequence design for expression engineering. Using our models, we study expression divergence under genetic drift and strong-selection weak-mutation regimes to find that regulatory evolution is rapid and subject to diminishing returns epistasis; that conflicting expression objectives in different environments constrain expression adaptation; and that stabilizing selection on gene expression leads to the moderation of regulatory complexity. We present an approach for using such models to detect signatures of selection on expression from natural variation in regulatory sequences and use it to discover an instance of convergent regulatory evolution. We assess mutational robustness, finding that regulatory mutation effect sizes follow a power law, characterize regulatory evolvability, visualize promoter fitness landscapes, discover evolvability archetypes and illustrate the mutational robustness of natural regulatory sequence populations. Our work provides a general framework for designing regulatory sequences and addressing fundamental questions in regulatory evolution.

Molecular Pathology of Lung Cancer

This overview of the molecular pathology of lung cancer includes a review of the most salient molecular alterations of the genome, transcriptome, and the epigenome. The insights provided by the growing use of next-generation sequencing (NGS) in lung cancer will be discussed, and interrelated concepts such as intertumor heterogeneity, intratumor heterogeneity, tumor mutational burden, and the advent of liquid biopsy will be explored. Moreover, this work describes how the evolving field of molecular pathology refines the understanding of different histologic phenotypes of non-small-cell lung cancer (NSCLC) and the underlying biology of small-cell lung cancer. This review will provide an appreciation for how ongoing scientific findings and technologic advances in molecular pathology are crucial for development of biomarkers, therapeutic agents, clinical trials, and ultimately improved patient care.

JASPAR 2022: the 9th release of the open-access database of transcription factor binding profiles

JASPAR (http://jaspar.genereg.net/) is an open-access database containing manually curated, non-redundant transcription factor (TF) binding profiles for TFs across six taxonomic groups. In this 9th release, we expanded the CORE collection with 341 new profiles (148 for plants, 101 for vertebrates, 85 for urochordates, and 7 for insects), which corresponds to a 19% expansion over the previous release. We added 298 new profiles to the Unvalidated collection when no orthogonal evidence was found in the literature. All the profiles were clustered to provide familial binding profiles for each taxonomic group. Moreover, we revised the structural classification of DNA binding domains to consider plant-specific TFs. This release introduces word clouds to represent the scientific knowledge associated with each TF. We updated the genome tracks of TFBSs predicted with JASPAR profiles in eight organisms; the human and mouse TFBS predictions can be visualized as native tracks in the UCSC Genome Browser. Finally, we provide a new tool to perform JASPAR TFBS enrichment analysis in user-provided genomic regions. All the data is accessible through the JASPAR website, its associated RESTful API, the R/Bioconductor data package, and a new Python package, pyJASPAR, that facilitates serverless access to the data.

Mapping single-cell data to reference atlases by transfer learning

Large single-cell atlases are now routinely generated to serve as references for analysis of smaller-scale studies. Yet learning from reference data is complicated by batch effects between datasets, limited availability of computational resources and sharing restrictions on raw data. Here we introduce a deep learning strategy for mapping query datasets on top of a reference called single-cell architectural surgery (scArches). scArches uses transfer learning and parameter optimization to enable efficient, decentralized, iterative reference building and contextualization of new datasets with existing references without sharing raw data. Using examples from mouse brain, pancreas, immune and whole-organism atlases, we show that scArches preserves biological state information while removing batch effects, despite using four orders of magnitude fewer parameters than de novo integration. scArches generalizes to multimodal reference mapping, allowing imputation of missing modalities. Finally, scArches retains coronavirus disease 2019 (COVID-19) disease variation when mapping to a healthy reference, enabling the discovery of disease-specific cell states. scArches will facilitate collaborative projects by enabling iterative construction, updating, sharing and efficient use of reference atlases.

Mapping single-cell data to reference atlases by transfer learning

Large single-cell atlases are now routinely generated to serve as references for analysis of smaller-scale studies. Yet learning from reference data is complicated by batch effects between datasets, limited availability of computational resources and sharing restrictions on raw data. Here we introduce a deep learning strategy for mapping query datasets on top of a reference called single-cell architectural surgery (scArches). scArches uses transfer learning and parameter optimization to enable efficient, decentralized, iterative reference building and contextualization of new datasets with existing references without sharing raw data. Using examples from mouse brain, pancreas, immune and whole-organism atlases, we show that scArches preserves biological state information while removing batch effects, despite using four orders of magnitude fewer parameters than de novo integration. scArches generalizes to multimodal reference mapping, allowing imputation of missing modalities. Finally, scArches retains coronavirus disease 2019 (COVID-19) disease variation when mapping to a healthy reference, enabling the discovery of disease-specific cell states. scArches will facilitate collaborative projects by enabling iterative construction, updating, sharing and efficient use of reference atlases.

OUP accepted manuscript

Motivation

Sequence models based on deep neural networks have achieved state-of-the-art performance on regulatory genomics prediction tasks, such as chromatin accessibility and transcription factor binding. But despite their high accuracy, their contributions to a mechanistic understanding of the biology of regulatory elements is often hindered by the complexity of the predictive model and thus poor interpretability of its decision boundaries. To address this, we introduce seqgra, a deep learning pipeline that incorporates the rule-based simulation of biological sequence data and the training and evaluation of models, whose decision boundaries mirror the rules from the simulation process.

Results

We show that seqgra can be used to (i) generate data under the assumption of a hypothesized model of genome regulation, (ii) identify neural network architectures capable of recovering the rules of said model and (iii) analyze a model’s predictive performance as a function of training set size and the complexity of the rules behind the simulated data.

Availability and implementation

The source code of the seqgra package is hosted on GitHub (https://github.com/gifford-lab/seqgra). seqgra is a pip-installable Python package. Extensive documentation can be found at https://kkrismer.github.io/seqgra.

Supplementary information

Supplementary data are available at Bioinformatics online.

A guide to machine learning for biologists

The expanding scale and inherent complexity of biological data have encouraged a growing use of machine learning in biology to build informative and predictive models of the underlying biological processes. All machine learning techniques fit models to data; however, the specific methods are quite varied and can at first glance seem bewildering. In this Review, we aim to provide readers with a gentle introduction to a few key machine learning techniques, including the most recently developed and widely used techniques involving deep neural networks. We describe how different techniques may be suited to specific types of biological data, and also discuss some best practices and points to consider when one is embarking on experiments involving machine learning. Some emerging directions in machine learning methodology are also discussed.

Categorizing metadata to help mobilize computable biomedical knowledge

INTRODUCTION

Computable biomedical knowledge artifacts (CBKs) are digital objects conveying biomedical knowledge in machine-interpretable structures. As more CBKs are produced and their complexity increases, the value obtained from sharing CBKs grows. Mobilizing CBKs and sharing them widely can only be achieved if the CBKs are findable, accessible, interoperable, reusable, and trustable (FAIR+T). To help mobilize CBKs, we describe our efforts to outline metadata categories to make CBKs FAIR+T.

METHODS

We examined the literature regarding metadata with the potential to make digital artifacts FAIR+T. We also examined metadata available online today for actual CBKs of 12 different types. With iterative refinement, we came to a consensus on key categories of metadata that, when taken together, can make CBKs FAIR+T. We use subject-predicate-object triples to more clearly differentiate metadata categories.

RESULTS

We defined 13 categories of CBK metadata most relevant to making CBKs FAIR+T. Eleven of these categories (type, domain, purpose, identification, location, CBK-to-CBK relationships, technical, authorization and rights management, provenance, evidential basis, and evidence from use metadata) are evident today where CBKs are stored online. Two additional categories (preservation and integrity metadata) were not evident in our examples. We provide a research agenda to guide further study and development of these and other metadata categories.

CONCLUSION

A wide variety of metadata elements in various categories is needed to make CBKs FAIR+T. More work is needed to develop a common framework for CBK metadata that can make CBKs FAIR+T for all stakeholders.

Multi-omic insights into Parkinson's Disease: From genetic associations to functional mechanisms

Genome-Wide Association Studies (GWAS) have elucidated the genetic components of Parkinson's Disease (PD). However, because the vast majority of GWAS association signals fall within non-coding regions, translating these results into an interpretable, mechanistic understanding of the disease etiology remains a major challenge in the field. In this review, we provide an overview of the approaches to prioritize putative causal variants and genes as well as summarise the primary findings of previous studies. We then discuss recent efforts to integrate multi-omics data to identify likely pathogenic cell types and biological pathways implicated in PD pathogenesis. We have compiled full summary statistics of cell-type, tissue, and phentoype enrichment analyses from multiple studies of PD GWAS and provided them in a standardized format as a resource for the research community (https://github.com/RajLabMSSM/PD_omics_review). Finally, we discuss the experimental, computational, and conceptual advances that will be necessary to fully elucidate the effects of functional variants and genes on cellular dysregulation and disease risk.

Gaining insight into metabolic diseases from human genetic discoveries

Human large-scale genetic association studies have identified sequence variations at thousands of genetic risk loci that are more common in patients with diverse metabolic disease compared with healthy controls. While these genetic associations have been replicated in multiple large cohorts and sometimes can explain up to 50% of heritability, the molecular and cellular mechanisms affected by common genetic variation associated with metabolic disease remains mostly unknown. A variety of new genome-wide data types, in conjunction with novel biostatistical and computational analytical methodologies and foundational experimental technologies, are paving the way for a principled approach to systematic variant-to-function (V2F) studies for metabolic diseases, turning associated regions into causal variants, cell types and states of action, effector genes, and cellular and physiological mechanisms. Identification of new target genes and cellular programs for metabolic risk loci will improve mechanistic understanding of disease biology and identification of novel therapeutic strategies.

Epitome: predicting epigenetic events in novel cell types with multi-cell deep ensemble learning

The accumulation of large epigenomics data consortiums provides us with the opportunity to extrapolate existing knowledge to new cell types and conditions. We propose Epitome, a deep neural network that learns similarities of chromatin accessibility between well characterized reference cell types and a query cellular context, and copies over signal of transcription factor binding and modification of histones from reference cell types when chromatin profiles are similar to the query. Epitome achieves state-of-the-art accuracy when predicting transcription factor binding sites on novel cellular contexts and can further improve predictions as more epigenetic signals are collected from both reference cell types and the query cellular context of interest.

Comparison of in silico strategies to prioritize rare genomic variants impacting RNA splicing for the diagnosis of genomic disorders

The development of computational methods to assess pathogenicity of pre-messenger RNA splicing variants is critical for diagnosis of human disease. We assessed the capability of eight algorithms, and a consensus approach, to prioritize 249 variants of uncertain significance (VUSs) that underwent splicing functional analyses. The capability of algorithms to differentiate VUSs away from the immediate splice site as being 'pathogenic' or 'benign' is likely to have substantial impact on diagnostic testing. We show that SpliceAI is the best single strategy in this regard, but that combined usage of tools using a weighted approach can increase accuracy further. We incorporated prioritization strategies alongside diagnostic testing for rare disorders. We show that 15% of 2783 referred individuals carry rare variants expected to impact splicing that were not initially identified as 'pathogenic' or 'likely pathogenic'; one in five of these cases could lead to new or refined diagnoses.

Global properties of regulatory sequences are predicted by transcription factor recognition mechanisms

Background

Mammalian genomes contain millions of putative regulatory sequences, which are delineated by binding of multiple transcription factors. The degree to which spacing and orientation constraints among transcription factor binding sites contribute to the recognition and identity of regulatory sequence is an unresolved but important question that impacts our understanding of genome function and evolution. Global mechanisms that underlie phenomena including the size of regulatory sequences, their uniqueness, and their evolutionary turnover remain poorly described.

Results

Here, we ask whether models incorporating different degrees of spacing and orientation constraints among transcription factor binding sites are broadly consistent with several global properties of regulatory sequence. These properties include length, sequence diversity, turnover rate, and dominance of specific TFs in regulatory site identity and cell type specification. Models with and without spacing and orientation constraints are generally consistent with all observed properties of regulatory sequence, and with regulatory sequences being fundamentally small (~ 1 nucleosome). Uniqueness of regulatory regions and their rapid evolutionary turnover are expected under all models examined. An intriguing issue we identify is that the complexity of eukaryotic regulatory sites must scale with the number of active transcription factors, in order to accomplish observed specificity.

Conclusions

Models of transcription factor binding with or without spacing and orientation constraints predict that regulatory sequences should be fundamentally short, unique, and turn over rapidly. We posit that the existence of master regulators may be, in part, a consequence of evolutionary pressure to limit the complexity and increase evolvability of regulatory sites.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13059-021-02503-y.

Reproducibility standards for machine learning in the life sciences

To make machine learning analyses in the life sciences more computationally reproducible, we propose standards based on data, model, and code publication, programming best practices, and workflow automation. By meeting these standards, the community of researchers applying machine learning methods in the life sciences can ensure that their analyses are worthy of trust.

this article has been peer reviewed.

Effective gene expression prediction from sequence by integrating long-range interactions

How noncoding DNA determines gene expression in different cell types is a major unsolved problem, and critical downstream applications in human genetics depend on improved solutions. Here, we report substantially improved gene expression prediction accuracy from DNA sequences through the use of a deep learning architecture, called Enformer, that is able to integrate information from long-range interactions (up to 100 kb away) in the genome. This improvement yielded more accurate variant effect predictions on gene expression for both natural genetic variants and saturation mutagenesis measured by massively parallel reporter assays. Furthermore, Enformer learned to predict enhancer-promoter interactions directly from the DNA sequence competitively with methods that take direct experimental data as input. We expect that these advances will enable more effective fine-mapping of human disease associations and provide a framework to interpret cis-regulatory evolution.

BioSeq-BLM: a platform for analyzing DNA, RNA and protein sequences based on biological language models

In order to uncover the meanings of ‘book of life’, 155 different biological language models (BLMs) for DNA, RNA and protein sequence analysis are discussed in this study, which are able to extract the linguistic properties of ‘book of life’. We also extend the BLMs into a system called BioSeq-BLM for automatically representing and analyzing the sequence data. Experimental results show that the predictors generated by BioSeq-BLM achieve comparable or even obviously better performance than the exiting state-of-the-art predictors published in literatures, indicating that BioSeq-BLM will provide new approaches for biological sequence analysis based on natural language processing technologies, and contribute to the development of this very important field. In order to help the readers to use BioSeq-BLM for their own experiments, the corresponding web server and stand-alone package are established and released, which can be freely accessed at http://bliulab.net/BioSeq-BLM/.

Biologically relevant transfer learning improves transcription factor binding prediction

BACKGROUND

Deep learning has proven to be a powerful technique for transcription factor (TF) binding prediction but requires large training datasets. Transfer learning can reduce the amount of data required for deep learning, while improving overall model performance, compared to training a separate model for each new task.

RESULTS

We assess a transfer learning strategy for TF binding prediction consisting of a pre-training step, wherein we train a multi-task model with multiple TFs, and a fine-tuning step, wherein we initialize single-task models for individual TFs with the weights learned by the multi-task model, after which the single-task models are trained at a lower learning rate. We corroborate that transfer learning improves model performance, especially if in the pre-training step the multi-task model is trained with biologically relevant TFs. We show the effectiveness of transfer learning for TFs with ~ 500 ChIP-seq peak regions. Using model interpretation techniques, we demonstrate that the features learned in the pre-training step are refined in the fine-tuning step to resemble the binding motif of the target TF (i.e., the recipient of transfer learning in the fine-tuning step). Moreover, pre-training with biologically relevant TFs allows single-task models in the fine-tuning step to learn useful features other than the motif of the target TF.

CONCLUSIONS

Our results confirm that transfer learning is a powerful technique for TF binding prediction.

Sfaira accelerates data and model reuse in single cell genomics

Single-cell RNA-seq datasets are often first analyzed independently without harnessing model fits from previous studies, and are then contextualized with public data sets, requiring time-consuming data wrangling. We address these issues with sfaira, a single-cell data zoo for public data sets paired with a model zoo for executable pre-trained models. The data zoo is designed to facilitate contribution of data sets using ontologies for metadata. We propose an adaption of cross-entropy loss for cell type classification tailored to datasets annotated at different levels of coarseness. We demonstrate the utility of sfaira by training models across anatomic data partitions on 8 million cells.

Decoding disease: from genomes to networks to phenotypes

Interpreting the effects of genetic variants is key to understanding individual susceptibility to disease and designing personalized therapeutic approaches. Modern experimental technologies are enabling the generation of massive compendia of human genome sequence data and associated molecular and phenotypic traits, together with genome-scale expression, epigenomics and other functional genomic data. Integrative computational models can leverage these data to understand variant impact, elucidate the effect of dysregulated genes on biological pathways in specific disease and tissue contexts, and interpret disease risk beyond what is feasible with experiments alone. In this Review, we discuss recent developments in machine learning algorithms for genome interpretation and for integrative molecular-level modelling of cells, tissues and organs relevant to disease. More specifically, we highlight existing methods and key challenges and opportunities in identifying specific disease-causing genetic variants and linking them to molecular pathways and, ultimately, to disease phenotypes.

Discovering differential genome sequence activity with interpretable and efficient deep learning

Discovering sequence features that differentially direct cells to alternate fates is key to understanding both cellular development and the consequences of disease related mutations. We introduce Expected Pattern Effect and Differential Expected Pattern Effect, two black-box methods that can interpret genome regulatory sequences for cell type-specific or condition specific patterns. We show that these methods identify relevant transcription factor motifs and spacings that are predictive of cell state-specific chromatin accessibility. Finally, we integrate these methods into framework that is readily accessible to non-experts and available for download as a binary or installed via PyPI or bioconda at https://cgs.csail.mit.edu/deepaccess-package/.