Biology subjects

Gerstein, M.

Publications and source records attributed to Gerstein, M..

At least 73 records · Page 4Linked to original sources

bioRxiv2020

Privacy-preserving genotype imputation with fully homomorphic encryption

Genotype imputation is the statistical inference of unknown genotypes using known population haplotype structures observed in large genomic datasets, such as HapMap and 1000 genomes project. Genotype imputation can help further our understanding of the relationships between genotypes and traits, and is extremely useful for analyses such as genome-wide association studies and expression quantitative loci inference. Increasing the number of genotyped genomes will increase the statistical power for inferring genotype-phenotype relationships, but the amount of data required and the compute-intense nature of the genotype imputation problem overwhelms servers. Hence, many institutions are moving towards outsourcing cloud services to scale up research in a cost effective manner. This raises privacy concerns, which we propose to address via homomorphic encryption. Homomorphic encryption is a type of encryption that allows data analysis on cipher texts, and would thereby avoid the decryption of private genotypes in the cloud. Here we develop an efficient, privacy-preserving genotype imputation algorithm, p-Impute, using homomorphic encryption. Our results showed that the performance of p-Impute is equivalent to the state-of-the-art plaintext solutions, achieving up to 99% micro area under curve score, and requiring a scalable amount of memory and computational time.

bioinformatics↗

bioRxiv2020

Bayesian Structural Time Series for Mobile Health and Sensor Data: A Flexible Modeling Framework for Evaluating Interventions

The development of mobile-health technology has the potential to revolutionize personalized medicine. Biomedical sensors (e.g. wearables) can assist with determining treatment plans for individuals, provide quantitative information to healthcare providers, and give objective measurements of health, leading to the goal of precise phenotypic correlates for genotypes. Even though treatments and interventions are becoming more specific and datasets more abundant, measuring the causal impact of health interventions requires careful considerations of complex covariate structures as well as knowledge of the temporal and spatial properties of the data. Thus, biomedical sensor data need to make use of specialized statistical models. Here, we show how the Bayesian structural time series framework, widely used in economics, can be applied to these data. We further show how this framework corrects for covariates to provide accurate assessments of interventions. Furthermore, it allows for a time-dependent confidence interval of impact, which is useful for considering individualized assessments of intervention efficacy. We provide a customized biomedical adaptor tool around a specific Google implementation of the Bayesian structural time series framework that uniformly processes, prepares, and registers diverse biomedical data. We apply the resulting software implementation to a structured set of examples in biomedicine to showcase the ability of the framework to evaluate interventions with varying levels of data richness and covariate complexity. In particular, we show how the framework is able to evaluate an exercise interventions effect on stabilizing blood glucose in a diabetes dataset. We also provide a future-anticipating illustration from a behavioral dataset showcasing how the framework integrates complex spatial covariates. Overall, we show the robustness of the Bayesian structural time series framework when applied to biomedical sensor data, highlighting its increasing value for current and future datasets.

bioinformatics↗

bioRxiv2020

Storing and analyzing a genome on a blockchain

The genomic characterization of individuals promises to be immensely useful for biomedical research and healthcare. However, a critical barrier to expanding personal genome sequencing is achieving secure, high-integrity storage of raw data. While cloud storage offers solutions to access such data from any place and device, the vulnerabilities of centralized storage in relation to security, data integrity, and robustness, such as single points of failure, have not yet been addressed. Blockchain is a potential alternative to these storage modes. However, storing large-scale data on blockchain can be challenging due to slow transaction speeds, the potential for chains to reach large sizes, and limitations on querying data stored on-chain. Currently, several genomic storage applications incorporate blockchain, but likely because of these challenges, many use blockchain only to facilitate and log data-access transactions, rather than to store raw genomic data on-chain. While this secures the process of data access, it does not secure the data itself, which is often stored off-chain (i.e. in a cloud or file-hosting services). Here, we developed a novel method of storing reference-aligned reads on-chain in a private blockchain network. We also developed tools for accessing and analyzing the on-chain data. We addressed the challenges of on-chain data storage by minimizing the data inserted to the chain using reference-based data compression techniques and by binning the on-chain data by genomic location to reduce retrieval times. Our tools provide open-source blockchain-based storage and access for advanced genomic analyses such as variant calling.

bioinformatics↗

bioRxiv2020

Predicting changes in protein thermostability upon point mutation with deep 3D convolutional neural networks

Predicting mutation-induced changes in protein thermodynamic stability ({triangleup}{triangleup}G) is of great interest in protein engineering, variant interpretation, and understanding protein biophysics. We introduce ThermoNet, a deep, 3D-convolutional neural network designed for structure-based prediction of {triangleup}{triangleup}Gs upon point mutation. To leverage the image-processing power inherent in convolutional neural networks, we treat protein structures as if they were multi-channel 3D images. In particular, the inputs to ThermoNet are uniformly constructed as multi-channel voxel grids based on biophysical properties derived from raw atom coordinates. We train and evaluate ThermoNet with a curated data set that accounts for protein homology and is balanced with direct and reverse mutations; this provides a framework for addressing biases that have likely influenced many previous {triangleup}{triangleup}G prediction methods. ThermoNet demonstrates performance comparable to the best available methods on the widely used Ssym test set. However, ThermoNet accurately predicts the effects of both stabilizing and destabilizing mutations, while most other methods exhibit a strong bias towards predicting destabilization. We further show that homology between Ssym and widely used training sets like S2648 and VariBench has likely led to overestimated performance in previous studies. Finally, we demonstrate the practical utility of ThermoNet in predicting the {triangleup}{triangleup}Gs for two clinically relevant proteins, p53 and myoglobin, and for pathogenic and benign missense variants from ClinVar. Overall, our results suggest that 3D convolutional neural networks can model the complex, non-linear interactions perturbed by mutations, directly from biophysical properties of atoms. Author SummaryThe thermodynamic stability of a protein, usually represented as the Gibbs free energy for the biophysical process of protein folding ({triangleup}G), is a fundamental thermodynamic quantity. Predicting mutation-induced changes in protein thermodynamic stability ({triangleup}{triangleup}G) is of great interest in protein engineering, variant interpretation, and understanding protein biophysics. However, predicting {triangleup}{triangleup}Gs in an accurate and unbiased manner has been a long-standing challenge in the field of computational biology. In this work, we introduce ThermoNet, a deep, 3D-convolutional neural network designed for structure-based {triangleup}{triangleup}G prediction. To leverage the image-processing power inherent in convolutional neural networks, we treat protein structures as if they were multi-channel 3D images. ThermoNet demonstrates performance comparable to the best available methods. However, ThermoNet accurately predicts the effects of both stabilizing and destabilizing mutations, while most other methods exhibit a strong bias towards predicting destabilization. We also demonstrate that the presence of homologous proteins in commonly used training and testing sets for {triangleup}{triangleup}G prediction methods has likely influenced previous performance estimates. Finally, we highlight the practical utility of ThermoNet by applying it to predicting the {triangleup}{triangleup}Gs for two clinically relevant proteins, p53 and myoglobin, and for pathogenic and benign missense variants from ClinVar.

bioinformatics↗

bioRxiv2020

Epigenome-based Splicing Prediction using a Recurrent Neural Network

Alternative RNA splicing provides an important means to expand metazoan transcriptome diversity. Contrary to what was accepted previously, splicing is now thought to predominantly take place during transcription. Motivated by emerging data showing the physical proximity of the spliceosome to Pol II, we surveyed the effect of epigenetic context on co-transcriptional splicing. In particular, we observed that splicing factors were not necessarily enriched at exon junctions and that most epigenetic signatures had a distinctly asymmetric profile around known splice sites. Given this, we tried to build an interpretable model that mimics the physical layout of splicing regulation where the chromatin context progressively changes as the Pol II moves along the guide DNA. We used a recurrent-neural-network architecture to predict the inclusion of a spliced exon based on adjacent epigenetic signals, and we showed that distinct spatio-temporal features of these signals were key determinants of model outcome, in addition to the actual nucleotide sequence of the guide DNA strand. After the model had been trained and tested (with >80% precision-recall curve metric), we explored the derived weights of the latent factors, finding they highlight the importance of the asymmetric time-direction of chromatin context during transcription. Author SummaryIn humans, only about 2% of the genome is comprised of so-called coding regions and can give rise to protein products. However, the human transcriptome is much more diverse than the number of genes found in these coding regions. Each gene can give rise to multiple transcripts through a process during transcription called alternative splicing. There is a limited understanding of the regulation of splicing and the underlying splicing code that determines cell-type-specific splicing. Here, we studied epigenetic features that characterize splicing regulation in humans using a recurrent neural network model. Unlike feedforward neural networks, this method contains an internal memory state that learns from spatiotemporal patterns - like the context in language - from a sequence of genomic and epigenetic information, making it better suited for characterizing splicing. We demonstrated that our method improves the prediction of spicing outcomes compared to previous methods. Furthermore, we applied our method to 49 cell types in ENCODE to investigate splicing regulation and found that not only spatial but also temporal epigenomic context can influence splicing regulation during transcription.

genomics↗

bioRxiv2019

To mock or not: a comprehensive comparison of mock IP and DNA input for ChIP-seq

Chromatin immunoprecipitation (IP) followed by sequencing (ChIP-seq) is the gold standard to detect genome-wide DNA-protein binding. The binding sites of transcription factors facilitate many biological studies. Of emerging concern is the abundance of spurious sites in ChIP-seq, which are mainly caused by uneven genomic sonication and nonspecific interactions between chromatin and antibody. A "mock" IP is designed to correct for both factors, whereas a DNA input control corrects only for uneven sonication. However, a mock IP is more susceptible to technical noise than a DNA input, and empirically, these two controls perform similarly for ChIP-seq. Therefore, DNA input is currently being used almost exclusively. With a large dataset, we demonstrate that using a DNA input control results in a definable set of spurious sites, and their abundance is tightly associated with the intrinsic properties of the ChIP-seq sample. For example, compared to human cell lines, samples such as human tissues and whole worm and fly have more accessible genomes, and thus have more spurious sites. The large and varying abundance of spurious sites may impede comparative studies across multiple samples. In contrast, using a mock IP as control substantially removes these spurious sites, resulting in high-quality binding sites and facilitating their comparability across samples. Although outperformed by mock IP, DNA input is still informative and has unique advantages. Therefore, we have developed a method to use both controls in combination to further improve binding site detection.

genomics↗

bioRxiv2019

Using Ethereum blockchain to store and query pharmacogenomics data via smart contracts

BackgroundWith the advent of precision medicine, pharmacogenomics data is becoming increasingly critical to patient care. These data describe the relationship between a particular variant in the genome and the response to a drug by the patient. As utilizing this kind of data becomes more integral to medical treatment decisions, appropriate storage and sharing of this data will be critical. A potential way of securely storing and sharing pharmacogenomics data is a smart contract with the Ethereum blockchain. This is an open-source blockchain platform for decentralized applications. A transaction-based, state machine, the "world" of Ethereum maintains user accounts and storage in a network state. Immutable pieces of code called "smart contracts" may be deployed to the Ethereum network and run on the Ethereum Virtual Machine when called by a user or other contract. The 2019 iDASH (Integrating Data for Analysis, Anonymization, and Sharing) competition for Secure Genome Analysis challenged participants to develop time- and space-efficient smart contracts to log and query gene-drug relationship data on the Ethereum blockchain. MethodsWe designed a smart contract to store and query pharmacogenomics data (gene-drug interaction data) in Ethereum using an index-based, multi-mapping approach allowing for time and space efficient storage and query. Our solution to the IDASH competition ranked in the top three at a workshop held in Bloomington, IN in October 2019. Although our solution performed well in the challenge, we wanted to improve its scalability and query efficiency. To that end, we developed an alternate "fastQuery" solution that stores pooled rather than raw data, allowing for significantly improved query time for 0-AND queries, and constant query time for 1- and 2-AND queries. ResultsWe tested the performance of both of our solutions in Truffle (v5.0.31) using datasets ranging from 100 to 1000 entries, and inserting data at 25, 50, 100, and 200 observations at a time. On a private, proof-of-authority test network, our challenge solution requires approximately 70 seconds, 500 MB of memory, and 80 MB of disk space to insert 1000 entries (200 at a time); and 400 ms and 5 MB of memory to query a two-AND query from 1000 entries. This solution exhibits constant memory for insertion and querying, and linear query time. Our alternate fastQuery solution requires approximately 60 seconds, 500 MB of memory, and 80 MB of disk space to insert 1000 entries (200 at a time); and 83 ms and 5 MB of memory to query a two-AND query from 1000 entries. This solution exhibits constant memory for insertion and querying, linear query time for 0-AND queries, and constant query time for 1- and 2-AND queries in a database of up to 1000 entries. ConclusionIn this study we showed that pharmacogenomics data can be stored and queried efficiently on the Ethereum blockchain. Our approach has the potential to be useful for a wide range of datasets in biomedical research; while we focused on gene-drug interaction data, our solution designs could be used to store a range of clinical trial data. Moreover, our solutions could be adapted to store and query data in any field where high-integrity data storage and efficient access is required.

bioinformatics↗

bioRxiv2019

TopicNet: a framework for measuring transcriptional regulatory network change

Next generation sequencing data highlights comprehensive and dynamic changes in the human gene regulatory network. Moreover, changes in regulatory network connectivity (network "rewiring") manifest different regulatory programs in multiple cellular states. However, due to the dense and noisy nature of the connectivity in regulatory networks, directly comparing the gains and losses of targets of key TFs is not that informative. Thus, here, we seek a abstracted lower-dimensional representation to understand the main features of network change. In particular, we propose a method called TopicNet that applies latent Dirichlet allocation (LDA) to extract meaningful functional topics for a collection of genes regulated by a TF. We then define a rewiring score to quantify the large-scale changes in the regulatory network in terms of topic change for a TF. Using this framework, we can pinpoint particular TFs that change greatly in network connectivity between different cellular states. This is particularly relevant in oncogenesis. Also, incorporating gene-expression data, we define a topic activity score that gives the degree that a topic is active in a particular cellular state. Furthermore, we show how activity differences can highlight differential survival in certain cancers.

bioinformatics↗

bioRxiv2019

Cyclic and Multilevel Causation in Evolutionary Processes

Many models of evolution are implicitly causal processes. Features such as causal feedback between evolutionary variables and evolutionary processes acting at multiple levels, though, mean that conventional causal models miss important phenomena. We develop here a general theoretical framework for analyzing evolutionary processes drawing on recent approaches to causal modeling developed in the machine-learning literature, which have extended Pearls do-calculus to incorporate cyclic causal interactions and multilevel causation. We also develop information-theoretic notions necessary to analyze causal information dynamics in our framework, introducing a causal generalization of the Partial Information Decomposition framework. We show how our causal framework helps to clarify conceptual issues in the contexts of complex trait analysis and cancer genetics, including assigning variation in an observed trait to genetic, epigenetic and environmental sources in the presence of epigenetic and environmental feedback processes, and variation in fitness to mutation processes in cancer using a multilevel causal model respectively, as well as relating causally-induced to observed variation in these variables via information theoretic bounds. In the process, we introduce a general class of multilevel causal evolutionary processes which connect evolutionary processes at multiple levels via coarse-graining relationships. Further, we show how a range of fitness models can be formulated in our framework, as well as a causal analog of Prices equation (generalizing the probabilistic Rice equation), clarifying the relationships between realized/probabilistic fitness and direct/indirect selection. Finally, we consider the potential relevance of our framework to foundational issues in biology and evolution, including supervenience, multilevel selection and individuality. Particularly, we argue that our class of multilevel causal evolutionary processes, in conjunction with a minimum description length principle, provides a conceptual framework in which identification of multiple levels of selection may be reduced to a model selection problem.

evolutionary biology↗

bioRxiv2019

FANCY: Fast Estimation of Privacy Risk in Functional Genomics Data

Functional genomics data is becoming clinically actionable, raising privacy concerns. However, quantifying the privacy leakage by genotyping is difficult due to the heterogeneous nature of sequencing techniques. Thus, we present FANCY, a tool that rapidly estimates number of leaking variants from raw RNA-Seq, ATAC-Seq and ChIP-Seq reads, without explicit genotyping. FANCY employs supervised regression using overall sequencing statistics as features and provides an estimate of the overall privacy risk before data release. FANCY can predict the cumulative number of leaking SNVs with a 0.95 average R2 for all independent test sets. We acknowledged the importance of accurate prediction even when the number of leaked variants is low, so we developed a special version of model, which can make predictions with higher accuracy for only a few leaking variants. A python and MATLAB implementation of FANCY, as well as custom scripts to generate the features can be found at https://github.com/gersteinlab/FANCY. We also provide jupyter notebooks so that users can optimize the parameters in the regression model based on their own data. An easy-to-use webserver that takes inputs and displays results can be found at fancy.gersteinlab.org.

bioinformatics↗

bioRxiv2019

Approaches for integrating heterogeneous RNA-seq data reveals cross-talk between microbes and genes in asthmatic patients

Sputum induction is a non-invasive method to evaluate the airway environment, particularly for asthma. RNA sequencing (RNAseq) can be used on sputum, but it can be challenging to interpret because sputum contains a complex and heterogeneous mixture of human cells and exogenous (microbial) material. In this study, we developed a methodology that integrates dimensionality reduction and statistical modeling to grapple with the heterogeneity. We use this to relate bulk RNAseq data from 115 asthmatic patients with clinical information, microscope images, and single-cell profiles. First, we mapped sputum RNAseq to human and exogenous sources. Next, we decomposed the human reads into cell-expression signatures and fractions of these in each sample; we validated the decomposition using targeted single-cell RNAseq and microscopy. We observed enrichment of immune-system cells (neutrophils, eosinophils, and mast cells) in severe asthmatics. Second, we inferred microbial abundances from the exogenous reads and then associated these with clinical variables -- e.g., Haemophilus was associated with increased white blood cell count and Candida, with worse lung function. Third, we applied a generative model, Latent Dirichlet allocation (LDA), to identify patterns of gene expression and microbial abundances and relate them to clinical data. Based on this, we developed a method called LDA-link that connects microbes to genes using reduced-dimensionality LDA topics. We found a number of known connections, e.g. between Haemophilus and the gene IL1B, which is highly expressed by mast cells. In addition, we identified novel connections, including Candida and the calcium-signaling gene CACNA1E, which is highly expressed by eosinophils. These results speak to the mechanism by which gene-microbe interactions contribute to asthma and define a strategy for making inferences in heterogeneous and noisy RNAseq datasets.

genomics↗

bioRxiv2019

Estimating growth patterns and driver effects in tumor evolution from individual samples

Evolving tumors accumulate thousands of mutations. Technological advances have enabled whole genome sequencing of these mutations in large cohorts, such as those from the Pancancer Analysis of Whole Genomes (PCAWG) Consortium. The resulting data explosion has led to many methods for detecting cancer drivers through mutational recurrence and deviation from background mutation rates. However, these methods require a large cohort and underperform when recurrence is low. An alternate approach involves harnessing the variant allele frequency (VAF) of mutations in the population of tumor cells in a single individual. Moreover, ultra-deep sequencing of tumors, which is now possible, allows for particularly accurate VAF measurements, and recent studies have begun to use these to determine evolutionary trajectories and quantify subclonal selection. Here, we developed a method that quantifies tumor growth and driver effects for individual samples based solely on the VAF spectrum. Drivers introduce a perturbation into this spectrum, and our method uses the frequency of \"hitchhiking\" mutations preceding a driver to measure this perturbation. Specifically, our method applies various growth models to identify periods of positive/negative growth, the genomic regions associated with them, and the presence and effect of putative drivers. To validate our method, we first used simulation models to successfully approximate the timing and size of a drivers effect. Then, we tested our method on 993 linear tumors (i.e. those with linear subclonal expansion, where each parent-subclone has one child) from the PCAWG Consortium and found that the identified periods of positive growth are associated with drivers previously highlighted via recurrence by the PCAWG consortium. Finally, we applied our method to an ultra-deep sequenced AML tumor and identified known cancer genes and additional driver candidates. In summary, our method presents opportunities for personalized diagnosis using deep sequenced whole genome data from an individual.

cancer biology↗

bioRxiv2019

SVFX: a machine-learning framework to quantify the pathogenicity of structural variants

A rapid decline in sequencing cost has made large-scale genome sequencing studies feasible. One of the fundamental goals of these studies is to catalog all pathogenic variants. Numerous methods and tools have been developed to interpret point mutations and small insertions and deletions. However, there is a lack of approaches for identifying pathogenic genomic structural variations (SVs). That said, SVs are known to play a crucial role in many diseases by altering the sequence and three-dimensional structure of the genome. Previous studies have suggested a complex interplay of genomic and epigenomic features in the emergence and distribution of SVs. However, the exact mechanism of pathogenesis for SVs in different diseases is not straightforward to decipher. Thus, we built an agnostic machine-learning-based workflow, called SVFX, to assign a \"pathogenicity score\" to somatic and germline SVs in various diseases. In particular, we generated somatic and germline training models, which included genomic, epigenomic, and conservation-based features for SV call sets in diseased and healthy individuals. We then applied SVFX to SVs in six different cancer cohorts and a cardiovascular disease (CVD) cohort. Overall, SVFX achieved high accuracy in identifying pathogenic SVs. Moreover, we found that predicted pathogenic SVs in cancer cohorts were enriched among known cancer genes and many cancer-related pathways (including Wnt signaling, Ras signaling, DNA repair, and ubiquitin-mediated proteolysis). Finally, we note that SVFX is flexible and can be easily extended to identify pathogenic SVs in additional disease cohorts.

bioinformatics↗

bioRxiv2019

STARRPeaker: Uniform processing and accurate identification of whole human STARR-seq active regions

BackgroundHigh-throughput reporter assays, such as self-transcribing active regulatory region sequencing (STARR-seq), allow for unbiased and quantitative assessment of enhancers at a genome-wide scale. Recent advances in STARR-seq technology have employed progressively more complex genomic libraries and increased sequencing depths, to assay larger sized regions, up to the entire human genome. These advances necessitate a reliable processing pipeline and peak-calling algorithm. ResultsMost STARR-seq studies have relied on chromatin immunoprecipitation sequencing (ChIP-seq) processing pipelines. However, there are key differences in STARR-seq versus ChIP-seq. First, STARR-seq uses transcribed RNA to measure the activity of an enhancer, making an accurate determination of the basal transcription rate important. Second, STARR-seq coverage is highly non-uniform, overdispersed, and often confounded by sequencing biases, such as GC content and mappability. Lastly, here, we observed a clear correlation between RNA thermodynamic stability and STARR-seq readout, suggesting that STARR-seq may be sensitive to RNA secondary structure and stability. Considering these findings, we developed a negative-binomial regression framework for uniformly processing STARR-seq data, called STARRPeaker. In support of this, we generated whole-genome STARR-seq data from the HepG2 and K562 human cell lines and applied STARRPeaker to call enhancers. ConclusionsWe show STARRPeaker can unbiasedly detect active enhancers from both captured and whole-genome STARR-seq data. Specifically, we report [~]33,000 and [~]20,000 candidate enhancers from HepG2 and K562, respectively. Moreover, we show that STARRPeaker outperforms other peak callers in terms of identifying known enhancers with fewer false positives. Overall, we demonstrate an optimized processing framework for STARR-seq experiments can identify putative enhancers while addressing potential confounders.

bioinformatics↗

bioRxiv2019

TeXP: Deconvolving the effects of pervasive and autonomous transcription of transposable elements

Long interspersed nuclear element 1 (LINE-1) is a primary source of genetic variation in humans and other mammals. Despite its importance, LINE-1 activity remains difficult to study because of its highly repetitive nature. Here, we developed and validated a method called TeXP to gauge LINE-1 activity accurately. TeXP builds mappability signatures from LINE-1 subfamilies to deconvolve the effect of pervasive transcription from autonomous LINE-1 activity. In particular, it apportions the multiple reads aligned to the many LINE-1 instances in the genome into these two categories. Using our method, we evaluated well-established cell lines, cell-line compartments and healthy tissues and found that the vast majority (91.7%) of transcriptome reads overlapping LINE-1 derive from pervasive transcription. We validated TeXP by independently estimating the levels of LINE-1 autonomous transcription using ddPCR, finding high concordance. Next, we applied our method to comprehensively measure LINE-1 activity across healthy somatic cells, while backing out the effect of pervasive transcription. Unexpectedly, we found that LINE-1 activity is present in many normal somatic cells. This finding contrasts with earlier studies showing that LINE-1 has limited activity in healthy somatic tissues, except for neuroprogenitor cells. Interestingly, we found that the amount of LINE-1 activity was associated with the with the amount of cell turnover, with tissues with low cell turnover rates (e.g. the adult central nervous system) showing lower LINE-1 activity. Altogether, our results show how accounting for pervasive transcription is critical to accurately quantify the activity of highly repetitive regions of the human genome.\n\nAuthor SummaryRepetitive sequences, such as LINEs, comprise more than half of the human genome. Due to their repetitive nature, LINEs are hard to grasp. In particular, we find that pervasive transcription is a major confounding factor in transcriptome data. We observe that, on average, more than 90% of LINE signal derives from pervasive transcription. To investigate this issue, we developed and validated a new method called TeXP. TeXP accounts and removes the effects of pervasive transcription when quantifying LINE activity. Our method uses the broad distribution of LINEs to estimate the effects of pervasive transcription. Using TeXP, we processed thousands of transcriptome datasets to uniformly, and unbiasedly measure LINE-1 activity across healthy somatic cells. By removing the pervasive transcription component, we find that (1) LINE-1 is broadly expressed in healthy somatic tissues; (2) Adult brain show small levels of LINE transcription and; (3) LINE-1 transcription level is correlated with tissue cell turnover. Our method thus offers insights into how repetitive sequences and influenced by pervasive transcription. Moreover, we uncover the activity of LINE-1 in somatic tissues at an unmatched scale.

bioinformatics↗

bioRxiv2018

Leveraging protein dynamics to identify cancer mutational hotspots in 3D-structures

Large-scale exome sequencing of tumors has enabled the identification of cancer drivers using recurrence and clustering-based approaches. Some of these methods also employ three-dimensional protein structures to identify mutational hotspots in cancer-associated genes. In determining such mutational clusters in structures, existing approaches overlook protein dynamics, despite the essential role of dynamics in protein functionality. In this work, we present a framework to identify driver genes using a dynamics-based search of mutational hotspot communities. After partitioning 3D structures into distinct communities of residues using anisotropic network models, we map variants onto the partitioned structures. We then search for signals of positive selection among these residue communities to identify putative drivers. We applied our method using the TCGA pan-cancer atlas missense mutation catalog. Overall, our analyses predict one or more mutational hotspots within the resolved structures of 434 genes. Ontological and pathway enrichment analyses implicate genes with predicted hotspots to be enriched in biological processes associated with tumor progression. Additionally, a comparison between our approach and existing hotspot detection methods that use structural data suggests that the inclusion of dynamics significantly increases the sensitivity of driver detection.

bioinformatics↗

bioRxiv2018

GRAM: A generalized model to predict the molecular effect of a non-coding variant in a cell-type specific manner

There has been much effort to prioritize genomic variants with respect to their impact on \"function\". However, function is often not precisely defined: Sometimes, it is the disease association of a variant; other times, it reflects a molecular effect on transcription or epigenetics. Here we coupled multiple genomic predictors to build GRAM, a generalized model, to predict a well-defined experimental target: the expression-modulating effect of a non-coding variant in a cell-specific manner. As a first step, we performed feature engineering: using a LASSO regularized linear model, we found transcription factor (TF) binding most predictive, especially for TFs that are hubs in the regulatory network; in contrast, evolutionary conservation, a popular feature in many other functional-impact predictors, has almost no contribution. Moreover, TF binding inferred from in vitro SELEX is as effective as that from in vivo ChIP-Seq. Second, we implemented GRAM integrating SELEX features and expression profiles. The program combines a universal regulatory score for a variant in a non-coding element with a modifier score reflecting the particular cell type. We benchmarked GRAM on a large-scale MPRA dataset in the GM12878 cell line, achieving a ROC score of [~]0.73; performance on the K562 cell line was similar. Finally, we evaluated the performance of GRAM on targeted regions using luciferase assays in MCF7 and K562 cell lines. We noted that changing the insertion position of the construct relative to the reporter gene gives very different results, highlighting the importance of carefully defining the functional target the model is predicting.\n\nAuthor SummaryNoncoding variants lie outside of protein-coding regions, and are found to have disease associations. However, knowledge on the molecular effect of these non-coding variants in a cell-specific context is very limited. Also, different output between multiple experiment platforms may introduce extra complexity in analyzing the molecular function of these variants. We developed GRAM, a generalized model to predict molecular effect of non-coding variants in multiple cell types for different experimental platforms. We first selected the most informative cell-independent SELEX transcription factor binding score on the variant locus as features and then combine cell-specific gene expression profile to build a multi-step prediction model. GRAM has been successfully tested on both MPRA and Luciferase assay, and on three different cell lines: GM12878, K562 and MCF7, shows high performance.

bioinformatics↗

bioRxiv2018

RADAR: annotation and prioritization of variants in the post-transcriptional regulome of RNA-binding proteins

RNA-binding proteins (RBPs) play key roles in post-transcriptional regulation and disease. Their binding sites cover more of the genome than coding exons; nevertheless, most noncoding variant-prioritization methods only focus on transcriptional regulation. Here, we integrate the portfolio of ENCODE-RBP experiments to develop RADAR, a variant-scoring framework. RADAR uses conservation, RNA structure, network centrality, and motifs to provide an overall impact score. Then it further incorporates tissue-specific inputs to highlight disease-specific variants. Our results demonstrate RADAR can successfully pinpoint variants, both somatic and germline, associated with RBP-function dysregulation, that cannot be found by most current prioritization methods, for example variants affecting splicing.

genomics↗