Joel T. Dudley

MEGA: A biologist-centric software for evolutionary analysis of DNA and protein sequences

The Molecular Evolutionary Genetics Analysis (MEGA) software is a desktop application designed for comparative analysis of homologous gene sequences either from multigene families or from different species with a special emphasis on inferring evolutionary relationships and patterns of DNA and protein evolution. In addition to the tools for statistical analysis of data, MEGA provides many convenient facilities for the assembly of sequence data sets from files or web-based repositories, and it includes tools for visual presentation of the results obtained in the form of interactive phylogenetic trees and evolutionary distance matrices. Here we discuss the motivation, design principles and priorities that have shaped the development of MEGA. We also discuss how MEGA might evolve in the future to assist researchers in their growing need to analyze large data set using new computational methods.

Personal Omics Profiling Reveals Dynamic Molecular and Medical Phenotypes

Personalized medicine is expected to benefit from combining genomic information with regular monitoring of physiological states by multiple high-throughput methods. Here, we present an integrative personal omics profile (iPOP), an analysis that combines genomic, transcriptomic, proteomic, metabolomic, and autoantibody profiles from a single individual over a 14 month period. Our iPOP analysis revealed various medical risks, including type 2 diabetes. It also uncovered extensive, dynamic changes in diverse molecular components and biological pathways across healthy and diseased conditions. Extremely high-coverage genomic and transcriptomic data, which provide the basis of our iPOP, revealed extensive heteroallelic changes during healthy and diseased states and an unexpected RNA editing mechanism. This study demonstrates that longitudinal iPOP can be used to interpret healthy and diseased states by connecting genomic information with additional dynamic omics activity.

Discovery and Preclinical Validation of Drug Indications Using Compendia of Public Gene Expression Data

A systematic computational method predicts new uses for existing drugs by integrating public gene expression signatures of drugs and diseases. Greening Drug Discovery Recycling is good for the environment—and for drug development too. Repurposing existing, approved drugs can speed their adoption in the clinic because they can often take advantage of the existing rigorous safety testing required by the Food and Drug Administration and other regulatory agencies. In a pair of papers, Sirota et al. and Dudley et al. examined publicly available gene expression data and determined the genes affected in 100 diseases and 164 drugs. By pairing drugs that correct abnormal gene expression in diseases, they confirm known effective drug-disease pairs and predict new indications for already approved agents. Experimental validation that an antiulcer drug and an antiepileptic can be reused for lung cancer and inflammatory bowel disease reinforces the promise of this approach. The authors scrutinized the data in Gene Expression Omnibus and identified a disease signature for 100 diseases, which they defined as the set of mRNAs that reliably increase or decrease in patients with that disease compared to normal individuals. They compared each of these disease signatures to each of the gene expression signatures for 164 drugs from the Connectivity Map, a collection of mRNA expression data from cultured human cells treated with bioactive small molecules that is maintained at the Broad Institute at Massachusetts Institute of Technology. A similarity score calculated by the authors for every possible pair of drug and disease ranged from +1 (a perfect correlation of signatures) to −1 (exactly opposite signatures). The investigators suggested that a similarity score of −1 would predict that the drug would ameliorate the abnormalities in the disease and thus be an effective therapy. This proved to be true for a number of drugs already on the market. The corticosteroid prednisolone, a common treatment for Crohn’s disease and ulcerative colitis, showed a strong similarity score for these two diseases. The histone deacetylase inhibitors trichostatin A, valproic acid, and vorinostat were predicted to work against brain tumors and other cancers (esophagus, lung, and colon), and there is experimental evidence that this is indeed the case. But in the ultimate test of method, the authors confirmed two new predictions in animal experiments: Cimetidine, an antiulcer drug, predicted by the authors to be effective against lung cancer, inhibited tumor cells in vitro and in vivo in mice. In addition, the antiepileptic topiramate, predicted to improve inflammatory bowel disease by similarity score, improved damage in colon tissue of rats treated with trinitrobenzenesulfonic acid, a model of the disease. These two drugs are therefore good candidates for recycling to treat two diseases in need of better therapies—lung cancer and inflammatory bowel disease—and we now have a way to mine available data for fast routes to new disease therapies. The application of established drug compounds to new therapeutic indications, known as drug repositioning, offers several advantages over traditional drug development, including reduced development costs and shorter paths to approval. Recent approaches to drug repositioning use high-throughput experimental approaches to assess a compound’s potential therapeutic qualities. Here, we present a systematic computational approach to predict novel therapeutic indications on the basis of comprehensive testing of molecular signatures in drug-disease pairs. We integrated gene expression measurements from 100 diseases and gene expression measurements on 164 drug compounds, yielding predicted therapeutic potentials for these drugs. We recovered many known drug and disease relationships using computationally derived therapeutic potentials and also predict many new indications for these 164 drugs. We experimentally validated a prediction for the antiulcer drug cimetidine as a candidate therapeutic in the treatment of lung adenocarcinoma, and demonstrate its efficacy both in vitro and in vivo using mouse xenograft models. This computational method provides a systematic approach for repositioning established drugs to treat a wide range of human diseases.

Exploiting drug–disease relationships for computational drug repositioning

Finding new uses for existing drugs, or drug repositioning, has been used as a strategy for decades to get drugs to more patients. As the ability to measure molecules in high-throughput ways has improved over the past decade, it is logical that such data might be useful for enabling drug repositioning through computational methods. Many computational predictions for new indications have been borne out in cellular model systems, though extensive animal model and clinical trial-based validation are still pending. In this review, we show that computational methods for drug repositioning can be classified in two axes: drug based, where discovery initiates from the chemical perspective, or disease based, where discovery initiates from the clinical perspective of disease or its pathology. Newer algorithms for computational drug repositioning will likely span these two axes, will take advantage of newer types of molecular measurements, and will certainly play a role in reducing the global burden of disease.

Deep Patient: An Unsupervised Representation to Predict the Future of Patients from the Electronic Health Records

Secondary use of electronic health records (EHRs) promises to advance clinical research and better inform clinical decision making. Challenges in summarizing and representing patient data prevent widespread practice of predictive modeling using EHRs. Here we present a novel unsupervised deep feature learning method to derive a general-purpose patient representation from EHR data that facilitates clinical predictive modeling. In particular, a three-layer stack of denoising autoencoders was used to capture hierarchical regularities and dependencies in the aggregated EHRs of about 700,000 patients from the Mount Sinai data warehouse. The result is a representation we name “deep patient”. We evaluated this representation as broadly predictive of health states by assessing the probability of patients to develop various diseases. We performed evaluation using 76,214 test patients comprising 78 diseases from diverse clinical domains and temporal windows. Our results significantly outperformed those achieved using representations based on raw EHR data and alternative feature learning strategies. Prediction performance for severe diabetes, schizophrenia, and various cancers were among the top performing. These findings indicate that deep learning applied to EHRs can derive patient representations that offer improved clinical predictions, and could provide a machine learning framework for augmenting clinical decision systems.

Phased Whole-Genome Genetic Risk in a Family Quartet Using a Major Allele Reference Sequence

Whole-genome sequencing harbors unprecedented potential for characterization of individual and family genetic variation. Here, we develop a novel synthetic human reference sequence that is ethnically concordant and use it for the analysis of genomes from a nuclear family with history of familial thrombophilia. We demonstrate that the use of the major allele reference sequence results in improved genotype accuracy for disease-associated variant loci. We infer recombination sites to the lowest median resolution demonstrated to date (<1,000 base pairs). We use family inheritance state analysis to control sequencing error and inform family-wide haplotype phasing, allowing quantification of genome-wide compound heterozygosity. We develop a sequence-based methodology for Human Leukocyte Antigen typing that contributes to disease risk prediction. Finally, we advance methods for analysis of disease and pharmacogenomic risk across the coding and non-coding genome that incorporate phased variant data. We show these methods are capable of identifying multigenic risk for inherited thrombophilia and informing the appropriate pharmacological therapy. These ethnicity-specific, family-based approaches to interpretation of genetic variation are emblematic of the next generation of genetic risk assessment using whole-genome sequencing.

Type 2 Diabetes Risk Alleles Demonstrate Extreme Directional Differentiation among Human Populations, Compared to Other Diseases

Many disease-susceptible SNPs exhibit significant disparity in ancestral and derived allele frequencies across worldwide populations. While previous studies have examined population differentiation of alleles at specific SNPs, global ethnic patterns of ensembles of disease risk alleles across human diseases are unexamined. To examine these patterns, we manually curated ethnic disease association data from 5,065 papers on human genetic studies representing 1,495 diseases, recording the precise risk alleles and their measured population frequencies and estimated effect sizes. We systematically compared the population frequencies of cross-ethnic risk alleles for each disease across 1,397 individuals from 11 HapMap populations, 1,064 individuals from 53 HGDP populations, and 49 individuals with whole-genome sequences from 10 populations. Type 2 diabetes (T2D) demonstrated extreme directional differentiation of risk allele frequencies across human populations, compared with null distributions of European-frequency matched control genomic alleles and risk alleles for other diseases. Most T2D risk alleles share a consistent pattern of decreasing frequencies along human migration into East Asia. Furthermore, we show that these patterns contribute to disparities in predicted genetic risk across 1,397 HapMap individuals, T2D genetic risk being consistently higher for individuals in the African populations and lower in the Asian populations, irrespective of the ethnicity considered in the initial discovery of risk alleles. We observed a similar pattern in the distribution of T2D Genetic Risk Scores, which are associated with an increased risk of developing diabetes in the Diabetes Prevention Program cohort, for the same individuals. This disparity may be attributable to the promotion of energy storage and usage appropriate to environments and inconsistent energy intake. Our results indicate that the differential frequencies of T2D risk alleles may contribute to the observed disparity in T2D incidence rates across ethnic populations.

Systematic functional regulatory assessment of disease-associated variants

Genome-wide association studies have discovered many genetic loci associated with disease traits, but the functional molecular basis of these associations is often unresolved. Genome-wide regulatory and gene expression profiles measured across individuals and diseases reflect downstream effects of genetic variation and may allow for functional assessment of disease-associated loci. Here, we present a unique approach for systematic integration of genetic disease associations, transcription factor binding among individuals, and gene expression data to assess the functional consequences of variants associated with hundreds of human diseases. In an analysis of genome-wide binding profiles of NFκB, we find that disease-associated SNPs are enriched in NFκB binding regions overall, and specifically for inflammatory-mediated diseases, such as asthma, rheumatoid arthritis, and coronary artery disease. Using genome-wide variation in transcription factor-binding data, we find that NFκB binding is often correlated with disease-associated variants in a genotype-specific and allele-specific manner. Furthermore, we show that this binding variation is often related to expression of nearby genes, which are also found to have altered expression in independent profiling of the variant-associated disease condition. Thus, using this integrative approach, we provide a unique means to assign putative function to many disease-associated SNPs.

Deep learning for healthcare: review, opportunities and challenges.

Gaining knowledge and actionable insights from complex, high-dimensional and heterogeneous biomedical data remains a key challenge in transforming health care. Various types of data have been emerging in modern biomedical research, including electronic health records, imaging, -omics, sensor data and text, which are complex, heterogeneous, poorly annotated and generally unstructured. Traditional data mining and statistical learning approaches typically need to first perform feature engineering to obtain effective and more robust features from those data, and then build prediction or clustering models on top of them. There are lots of challenges on both steps in a scenario of complicated data and lacking of sufficient domain knowledge. The latest advances in deep learning technologies provide new effective paradigms to obtain end-to-end learning models from complex data. In this article, we review the recent literature on applying deep learning technologies to advance the health care domain. Based on the analyzed work, we suggest that deep learning approaches could be the vehicle for translating big biomedical data into improved human health. However, we also note limitations and needs for improved methods development and applications, especially in terms of ease-of-understanding for domain experts and citizen scientists. We discuss such challenges and suggest developing holistic and meaningful interpretable architectures to bridge deep learning models and human interpretability.

Identification of discriminating biomarkers for human disease using integrative network biology.

There is a strong clinical imperative to identify discerning molecular biomarkers of disease to inform diagnosis, prognosis, and treatment. Ideally, such biomarkers would be drawn from peripheral sources non-invasively to reduce costs and lower potential for complication. Advances in high-throughput genomics and proteomics have vastly increased the space of prospective molecular biomarkers. Consequently, the elucidation of molecular biomarkers of clinical importance often entails a genome- or proteome-wide search for candidates. Here we present a novel framework for the identification of disease-specific protein biomarkers through the integration of biofluid proteomes and inter-disease genomic relationships using a network paradigm. We created a blood plasma biomarker network by linking expression-based genomic profiles from 136 diseases to 1,028 detectable blood plasma proteins. We also created a urine biomarker network by linking genomic profiles from 127 diseases to 577 proteins detectable in urine. Through analysis of these molecular biomarker networks, we find that the majority (> 80%) of putative protein biomarkers are linked to multiple disease conditions. Thus, prospective disease-specific protein biomarkers are found in only a small subset of the biofluids proteomes. These findings illustrate the importance of considering shared molecular pathology across diseases when evaluating biomarker specificity. The proposed framework is amenable to integration with complimentary network models of biology, which could further constrain the biomarker candidate space, and establish a role for the understanding of multi-scale, inter-disease genomic relationships in biomarker discovery.