The fast replication rate and lack of repair mechanisms of human immunodeficiency virus (HIV) contribute to its high mutation frequency, with some mutations resulting in the evolution of resistance to antiretroviral therapies (ART). As such, studying HIV drug resistance allows for real-time evaluation of evolutionary mechanisms. Characterizing the biological process of drug resistance is also critically important for sustained effectiveness of ART. Investigating the link between “black box” deep learning methods applied to this problem and evolutionary principles governing drug resistance has been overlooked to date. Here, we utilized publicly available HIV-1 sequence data and drug resistance assay results for 18 ART drugs to evaluate the performance of three architectures (multilayer perceptron, bidirectional recurrent neural network, and convolutional neural network) for drug resistance prediction, jointly with biological analysis. We identified convolutional neural networks as the best performing architecture and displayed a correspondence between the importance of biologically relevant features in the classifier and overall performance. Our results suggest that the high classification performance of deep learning models is indeed dependent on drug resistance mutations (DRMs). These models heavily weighted several features that are not known DRM locations, indicating the utility of model interpretability to address causal relationships in viral genotype-phenotype data.

Human endogenous retroviruses (HERV) have been implicated in a variety of diseases including cancers. Recent research implicates HERVs in epigenetic gene regulation. Here we utilize a recently developed bioinformatics tool for identifying HERV expression at the locus-specific level to identify differential expression of HERVs in matched tumor-normal RNA-sequencing (RNA-seq) data from The Cancer Genome Atlas. Data from 52 prostate cancer, 111 breast cancer, and 24 colon cancer cases were analyzed. Locus-specific analysis identified active HERV elements and differentially expressed HERVs in prostate cancer, breast cancer, and colon cancer. In addition, differentially expressed host genes were identified across prostate, breast, and colon cancer datasets, respectively, including several involved in demethylation and antiviral response pathways, supporting previous findings regarding the pathogenic mechanisms of HERVs. A majority of differentially expressed HERVs intersected protein coding genes or lncRNAs in each dataset, and a subset of differentially expressed HERVs intersected differentially expressed genes in prostate, breast, and colon cancers, providing evidence towards regulatory function. Finally, patterns in HERV expression were identified in multiple cancer types, with 155 HERVs differentially expressed in all three cancer types. This analysis extends previous results identifying HERV transcription in cancer RNA-seq datasets to a locus-specific level, and in doing so provides a foundation for future studies investigating the functional role of HERV in cancers and identifies a number of novel targets for cancer biomarkers and immunotherapy. Significance: Expressed human endogenous retroviruses are mapped at locus-specific resolution and linked to specific pathways to identify potential biomarkers and therapeutic targets in prostate, breast, and colon cancers.

Theoretical population genetics has long studied the arrival and geographic spread of adaptive variants through the analysis of mathematical models of dispersal and natural selection. These models take on a renewed interest in the context of the COVID-19 pandemic, especially given the consequences that novel adaptive variants have had on the course of the pandemic as they have spread through global populations. Here, we review theoretical models for the spatial spread of adaptive variants and identify areas to be improved in future work, toward a better understanding of variants of concern in Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) evolution and other contemporary applications. As we describe, characteristics of pandemics such as COVID-19—such as the impact of long-distance travel patterns and the overdispersion of lineages due to superspreading events—suggest new directions for improving upon existing population genetic models.