Epistasis and entrenchment of drug resistance in HIV-1 subtype B

Changing selection on amino acid substitutions in Gag protein between major HIV-1 subtypes

Amino acid preferences at a protein site depend on the role of this site in protein function and structure as well as on external constraints. All these factors can change in the course of evolution, making amino acid propensities of a site time-dependent. When viral subtypes divergently evolve in different host subpopulations, such changes may depend on genetic, medical, and sociocultural differences between these subpopulations. Here, using our previously developed phylogenetic approach, we describe sixty-nine amino acid sites of the Gag protein of human immunodeficiency virus type 1 (HIV-1) where amino acids have different impact on viral fitness in six major subtypes of the type M. These changes in preferences trigger adaptive evolution; indeed, 32 (46 per cent) of these sites experienced strong positive selection at least in one of the subtypes. At some of the sites, changes in amino acid preferences may be associated with differences in immune escape between subtypes. The prevalence of an amino acid in a protein site within a subtype is only a poor predictor for whether this amino acid is preferred in this subtype according to the phylogenetic analysis. Therefore, attempts to identify the factors of viral evolution from comparative genomics data should integrate across multiple sources of information.

Kinetic coevolutionary models predict the temporal emergence of HIV-1 resistance mutations under drug selection pressure

Drug resistance in HIV type 1 (HIV-1) is a pervasive problem that affects the lives of millions of people worldwide. Although records of drug-resistant mutations (DRMs) have been extensively tabulated within public repositories, our understanding of the evolutionary kinetics of DRMs and how they evolve together remains limited. Epistasis, the interaction between a DRM and other residues in HIV-1 protein sequences, is key to the temporal evolution of drug resistance. We use a Potts sequence-covariation statistical-energy model of HIV-1 protein fitness under drug selection pressure, which captures epistatic interactions between all positions, combined with kinetic Monte-Carlo simulations of sequence evolutionary trajectories, to explore the acquisition of DRMs as they arise in an ensemble of drug-naive patient protein sequences. We follow the time course of 52 DRMs in the enzymes protease, RT, and integrase, the primary targets of antiretroviral therapy. The rates at which DRMs emerge are highly correlated with their observed acquisition rates reported in the literature when drug pressure is applied. This result highlights the central role of epistasis in determining the kinetics governing DRM emergence. Whereas rapidly acquired DRMs begin to accumulate as soon as drug pressure is applied, slowly acquired DRMs are contingent on accessory mutations that appear only after prolonged drug pressure. We provide a foundation for using computational methods to determine the temporal evolution of drug resistance using Potts statistical potentials, which can be used to gain mechanistic insights into drug resistance pathways in HIV-1 and other infectious agents.

In vivo functional phenotypes from a computational epistatic model of evolution

Computational models of evolution are valuable for understanding the dynamics of sequence variation, to infer phylogenetic relationships or potential evolutionary pathways and for biomedical and industrial applications. Despite these benefits, few have validated their propensities to generate outputs with in vivo functionality, which would enhance their value as accurate and interpretable evolutionary algorithms. We demonstrate the power of epistasis inferred from natural protein families to evolve sequence variants in an algorithm we developed called sequence evolution with epistatic contributions (SEEC). Utilizing the Hamiltonian of the joint probability of sequences in the family as fitness metric, we sampled and experimentally tested for in vivo [Formula: see text]-lactamase activity in Escherichia coli TEM-1 variants. These evolved proteins can have dozens of mutations dispersed across the structure while preserving sites essential for both catalysis and interactions. Remarkably, these variants retain family-like functionality while being more active than their wild-type predecessor. We found that depending on the inference method used to generate the epistatic constraints, different parameters simulate diverse selection strengths. Under weaker selection, local Hamiltonian fluctuations reliably predict relative changes to variant fitness, recapitulating neutral evolution. SEEC has the potential to explore the dynamics of neofunctionalization, characterize viral fitness landscapes, and facilitate vaccine development.

Revolutionizing HIV-1 Viral Load Monitoring in India: The Potential of Dried Blood Spot Analysis for Expanding Access and Improving Care

India continues to grapple with a significant burden of HIV infections. Despite notable progress in prevention and treatment efforts, multiple challenges, such as high-risk populations, inadequate testing facilities, and limited access to healthcare in remote areas, persist. Though the Government of India offers HIV-1 plasma viral load testing at various medical centers, aiding treatment decisions and monitoring antiretroviral therapy effectiveness, enhancing care for individuals living with HIV under the National AIDS Control Program (NACP), the nation's large population and diverse demographics further complicate its outreach and response. Hence, strategic interventions and alternative methods of testing remain crucial to curbing HIV transmission and improving the quality of life for those affected. Dried blood spot (DBS) sampling has emerged as a convenient and cost-effective alternative for HIV-1 viral load testing, revolutionizing the landscape of diagnostic and monitoring strategies for HIV infection. Though the plasma-based viral load remains the gold standard for monitoring HIV-1, DBS-based HIV-1 viral load testing holds immense promise for improving access to care, particularly in resource-limited settings where traditional plasma-based methods may be logistically challenging. DBS entails the collection of a small volume of blood onto filter paper, followed by drying and storage. This approach offers numerous advantages, including simplified sample collection, transportation, and storage, reducing the need for cold-chain logistics. Recent studies have demonstrated the feasibility and accuracy of DBS-based HIV-1 viral load testing, revealing a strong correlation between DBS and plasma measurements. Its implementation can enhance the early detection of treatment failure, guide therapeutic decisions, and ultimately contribute to better clinical outcomes for HIV-infected individuals. Hence, this review explores the principles, advancements, feasibility, and implications of DBS-based HIV-1 viral load testing.

Direct-acting antiviral resistance of Hepatitis C virus is promoted by epistasis

Direct-acting antiviral agents (DAAs) provide efficacious therapeutic treatments for chronic Hepatitis C virus (HCV) infection. However, emergence of drug resistance mutations (DRMs) can greatly affect treatment outcomes and impede virological cure. While multiple DRMs have been observed for all currently used DAAs, the evolutionary determinants of such mutations are not currently well understood. Here, by considering DAAs targeting the nonstructural 3 (NS3) protein of HCV, we present results suggesting that epistasis plays an important role in the evolution of DRMs. Employing a sequence-based fitness landscape model whose predictions correlate highly with experimental data, we identify specific DRMs that are associated with strong epistatic interactions, and these are found to be enriched in multiple NS3-specific DAAs. Evolutionary modelling further supports that the identified DRMs involve compensatory mutational interactions that facilitate relatively easy escape from drug-induced selection pressures. Our results indicate that accounting for epistasis is important for designing future HCV NS3-targeting DAAs.

Mathematical Models of HIV-1 Dynamics, Transcription, and Latency

HIV-1 latency is a major barrier to curing infections with antiretroviral therapy and, consequently, to eliminating the disease globally. The establishment, maintenance, and potential clearance of latent infection are complex dynamic processes and can be best described with the help of mathematical models followed by experimental validation. Here, we review the use of viral dynamics models for HIV-1, with a focus on applications to the latent reservoir. Such models have been used to explain the multi-phasic decay of viral load during antiretroviral therapy, the early seeding of the latent reservoir during acute infection and the limited inflow during treatment, the dynamics of viral blips, and the phenomenon of post-treatment control. Finally, we discuss that mathematical models have been used to predict the efficacy of potential HIV-1 cure strategies, such as latency-reversing agents, early treatment initiation, or gene therapies, and to provide guidance for designing trials of these novel interventions.

Mechanisms of HIV-1 integrase resistance to dolutegravir and potent inhibition of drug-resistant variants

HIV-1 infection depends on the integration of viral DNA into host chromatin. Integration is mediated by the viral enzyme integrase and is blocked by integrase strand transfer inhibitors (INSTIs), first-line antiretroviral therapeutics widely used in the clinic. Resistance to even the best INSTIs is a problem, and the mechanisms of resistance are poorly understood. Here, we analyze combinations of the mutations E138K, G140A/S, and Q148H/K/R, which confer resistance to INSTIs. The investigational drug 4d more effectively inhibited the mutants compared with the approved drug Dolutegravir (DTG). We present 11 new cryo-EM structures of drug-resistant HIV-1 intasomes bound to DTG or 4d, with better than 3-Å resolution. These structures, complemented with free energy simulations, virology, and enzymology, explain the mechanisms of DTG resistance involving E138K + G140A/S + Q148H/K/R and show why 4d maintains potency better than DTG. These data establish a foundation for further development of INSTIs that potently inhibit resistant forms in integrase.

A personalized pharmaco-epistatic network model of precision medicine

Precision medicine, the utilization of targeted treatments to address an individual's disease, relies on knowledge about the genetic cause of that individual's drug response. Here, we present a functional graph (FunGraph) theory to chart comprehensive pharmacogenetic architecture for each and every patient. FunGraph is the combination of functional mapping - a dynamic model for genetic mapping and evolutionary game theory guiding interactive strategies. It coalesces all pharmacogenetic factors into multilayer and multiplex networks that fully capture bidirectional, signed and weighted epistasis. It can visualize and interrogate how epistasis moves in the cell and how this movement leads to patient- and context-specific genetic architecture in response to organismic physiology. We discuss the future implementation of FunGraph to achieve precision medicine. Teaser: We present a functional graph (FunGraph) theory to draw a complete picture of pharmacogenetic architecture underlying interindividual variability in drug response. FunGraph can characterize how each gene acts and interacts with every other gene to mediate therapeutic response.

Machine learning aided multiscale modelling of the HIV-1 infection in the presence of NRTI therapy

Human Immunodeficiency Virus (HIV) is one of the most common chronic infectious diseases in humans. Extending the expected lifetime of patients depends on the use of optimal antiretroviral therapies. Emergence of the drug-resistant strains can reduce the effectiveness of treatments and lead to Acquired Immunodeficiency Syndrome (AIDS), even with antiretroviral therapy. Investigating the genotype-phenotype relationship is a crucial process for optimizing the therapy protocols of the patients. Here, a mathematical modelling framework is proposed to address the impact of existing mutations, timing of initiation, and adherence levels of nucleotide reverse transcriptase inhibitors (NRTIs) on the evolutionary dynamics of the virus strains. For the first time, the existing Stanford HIV drug resistance data have been combined with a multi-strain within-host ordinary differential equation (ODE) model to track the dynamics of the most common NRTI-resistant strains. Overall, the D4T-3TC, D4T-AZT and TDF-D4T drug combinations have been shown to provide higher success rates in preventing treatment failure and further drug resistance. The results are in line with the genotype-phenotype data and pharmacokinetic parameters of the NRTI inhibitors. Moreover, we show that the undetectable mutant strains at the diagnosis have a significant effect on the success/failure rates of the NRTI treatments. Predictions on undetectable strains through our multi-strain within-host model yielded the possible role of viral evolution on the treatment outcomes. It has been recognized that the improvement of multi-scale models can contribute to the understanding of the evolutionary dynamics, and treatment options, and potentially increase the reliability of genotype-phenotype models.

Efficacy of Dolutegravir versus Darunavir in Antiretroviral First-Line Regimens According to Resistance Mutations and Viral Subtype

Background: Dolutegravir (DTG)-based first-line regimens have shown superior efficacy versus darunavir (DRV)-based ones in randomized trials. We compared these two strategies in clinical practice, particularly considering the role of pre-treatment drug resistance mutations (DRMs) and of the HIV-1 subtype. Materials and methods: The multicenter Antiretroviral Resistance Cohort Analysis (ARCA) database was queried to identify HIV-1-positive patients starting a first-line therapy with 2NRTIs plus either DTG or DRV between 2013 and 2019. Only adult (≥18 years) patients with a genotypic resistance test (GRT) prior to therapy and with HIV-1 RNA ≥1000 copies/mL were selected. Through multivariable Cox regressions, we compared DTG- versus DRV-based regimens in the time to virological failure (VF) stratifying for pre-treatment DRMs and the viral subtype. Results: A total of 649 patients was enrolled, with 359 (55.3%) and 290 (44.7) starting DRV and DTG, respectively. In 11 months of median follow-up time, there were 41 VFs (8.4 in 100 patient-years follow-up, PYFU) and 15 VFs (5.3 per 100 PYFU) in the DRV and DTG groups, respectively. Compared with a fully active DTG-based regimen, the risk of VF was higher with DRV (aHR 2.33; p = 0.016), and with DTG-based regimens with pre-treatment DRMs to the backbone (aHR 17.27; p = 0.001), after adjusting for age, gender, baseline CD4 count and HIV-RNA, concurrent AIDS-defining event and months since HIV diagnosis. Compared with patients harboring a B viral subtype and treated with a DTG-based regimen, patients on DRV had an increased risk of VF, both in subtype B (aHR 3.35; p = 0.011), C (aHR 8.10; p = 0.005), CRF02-AG (aHR 5.59; p = 0.006) and G (aHR 13.90; p < 0.001); DTG also demonstrated a reduced efficacy in subtypes C (versus B, aHR 10.24; p = 0.035) and CRF01-AE (versus B; aHR 10.65; p = 0.035). Higher baseline HIV-RNA and a longer time since HIV diagnosis also predicted VF. Conclusions: In line with randomized trials, DTG-based first-line regimens showed an overall superior efficacy compared with DRV-based regimens. GRT may still play a role in identifying patients more at risk of VF and in guiding the choice of an antiretroviral backbone.

Contingency and Entrenchment of Drug-Resistance Mutations in HIV Viral Proteins

The ability of HIV-1 to rapidly mutate leads to antiretroviral therapy (ART) failure among infected patients. Drug-resistance mutations (DRMs), which cause a fitness penalty to intrinsic viral fitness, are compensated by accessory mutations with favorable epistatic interactions which cause an evolutionary trapping effect, but the kinetics of this overall process has not been well characterized. Here, using a Potts Hamiltonian model describing epistasis combined with kinetic Monte Carlo simulations of evolutionary trajectories, we explore how epistasis modulates the evolutionary dynamics of HIV DRMs. We show how the occurrence of a drug-resistance mutation is contingent on favorable epistatic interactions with many other residues of the sequence background and that subsequent mutations entrench DRMs. We measure the time-autocorrelation of fluctuations in the likelihood of DRMs due to epistatic coupling with the sequence background, which reveals the presence of two evolutionary processes controlling DRM kinetics with two distinct time scales. Further analysis of waiting times for the evolutionary trapping effect to reverse reveals that the sequences which entrench (trap) a DRM are responsible for the slower time scale. We also quantify the overall strength of epistatic effects on the evolutionary kinetics for different mutations and show these are much larger for DRM positions than polymorphic positions, and we also show that trapping of a DRM is often caused by the collective effect of many accessory mutations, rather than a few strongly coupled ones, suggesting the importance of multiresidue sequence variations in HIV evolution. The analysis presented here provides a framework to explore the kinetic pathways through which viral proteins like HIV evolve under drug-selection pressure.

Computational compensatory mutation discovery approach: Predicting a PARP1 variant rescue mutation

The prediction of protein mutations that affect function may be exploited for multiple uses. In the context of disease variants, the prediction of compensatory mutations that reestablish functional phenotypes could aid in the development of genetic therapies. In this work, we present an integrated approach that combines coevolutionary analysis and molecular dynamics (MD) simulations to discover functional compensatory mutations. This approach is employed to investigate possible rescue mutations of a poly(ADP-ribose) polymerase 1 (PARP1) variant, PARP1 V762A, associated with lung cancer and follicular lymphoma. MD simulations show PARP1 V762A exhibits noticeable changes in structural and dynamical behavior compared with wild-type (WT) PARP1. Our integrated approach predicts A755E as a possible compensatory mutation based on coevolutionary information, and molecular simulations indicate that the PARP1 A755E/V762A double mutant exhibits similar structural and dynamical behavior to WT PARP1. Our methodology can be broadly applied to a large number of systems where single-nucleotide polymorphisms have been identified as connected to disease and can shed light on the biophysical effects of such changes as well as provide a way to discover potential mutants that could restore WT-like functionality. This can, in turn, be further utilized in the design of molecular therapeutics that aim to mimic such compensatory effect.

Approved HIV reverse transcriptase inhibitors in the past decade

HIV reverse transcriptase (RT) inhibitors are the important components of highly active antiretroviral therapies (HAARTs) for anti-HIV treatment and pre-exposure prophylaxis in clinical practice. Many RT inhibitors and their combination regimens have been approved in the past ten years, but a review on their drug discovery, pharmacology, and clinical efficacy is lacking. Here, we provide a comprehensive review of RT inhibitors (tenofovir alafenamide, rilpivirine, doravirine, dapivirine, azvudine and elsulfavirine) approved in the past decade, regarding their drug discovery, pharmacology, and clinical efficacy in randomized controlled trials. Novel RT inhibitors such as islatravir, MK-8504, MK-8507, MK8583, IQP-0528, and MIV-150 will be also highlighted. Future development may focus on the new generation of novel antiretroviral inhibitors with higher bioavailability, longer elimination half-life, more favorable side-effect profiles, fewer drug-drug interactions, and higher activities against circulating drug-resistant strains.

Limits to detecting epistasis in the fitness landscape of HIV

The rapid evolution of HIV is constrained by interactions between mutations which affect viral fitness. In this work, we explore the role of epistasis in determining the mutational fitness landscape of HIV for multiple drug target proteins, including Protease, Reverse Transcriptase, and Integrase. Epistatic interactions between residues modulate the mutation patterns involved in drug resistance, with unambiguous signatures of epistasis best seen in the comparison of the Potts model predicted and experimental HIV sequence “prevalences” expressed as higher-order marginals (beyond triplets) of the sequence probability distribution. In contrast, experimental measures of fitness such as viral replicative capacities generally probe fitness effects of point mutations in a single background, providing weak evidence for epistasis in viral systems. The detectable effects of epistasis are obscured by higher evolutionary conservation at sites. While double mutant cycles in principle, provide one of the best ways to probe epistatic interactions experimentally without reference to a particular background, we show that the analysis is complicated by the small dynamic range of measurements. Overall, we show that global pairwise interaction Potts models are necessary for predicting the mutational landscape of viral proteins.

Q-omics: Smart Software for Assisting Oncology and Cancer Research

The rapid increase in collateral omics and phenotypic data has enabled data-driven studies for the fast discovery of cancer targets and biomarkers. Thus, it is necessary to develop convenient tools for general oncologists and cancer scientists to carry out customized data mining without computational expertise. For this purpose, we developed innovative software that enables user-driven analyses assisted by knowledge-based smart systems. Publicly available data on mutations, gene expression, patient survival, immune score, drug screening and RNAi screening were integrated from the TCGA, GDSC, CCLE, NCI, and DepMap databases. The optimal selection of samples and other filtering options were guided by the smart function of the software for data mining and visualization on Kaplan-Meier plots, box plots and scatter plots of publication quality. We implemented unique algorithms for both data mining and visualization, thus simplifying and accelerating user-driven discovery activities on large multiomics datasets. The present Q-omics software program (v0.95) is available at http://qomics.sookmyung.ac.kr.

The generative capacity of probabilistic protein sequence models

Potts models and variational autoencoders (VAEs) have recently gained popularity as generative protein sequence models (GPSMs) to explore fitness landscapes and predict mutation effects. Despite encouraging results, current model evaluation metrics leave unclear whether GPSMs faithfully reproduce the complex multi-residue mutational patterns observed in natural sequences due to epistasis. Here, we develop a set of sequence statistics to assess the "generative capacity" of three current GPSMs: the pairwise Potts Hamiltonian, the VAE, and the site-independent model. We show that the Potts model's generative capacity is largest, as the higher-order mutational statistics generated by the model agree with those observed for natural sequences, while the VAE's lies between the Potts and site-independent models. Importantly, our work provides a new framework for evaluating and interpreting GPSM accuracy which emphasizes the role of higher-order covariation and epistasis, with broader implications for probabilistic sequence models in general.

Deep Mutational Scanning of Viral Glycoproteins and Their Host Receptors

Deep mutational scanning or deep mutagenesis is a powerful tool for understanding the sequence diversity available to viruses for adaptation in a laboratory setting. It generally involves tracking an in vitro selection of protein sequence variants with deep sequencing to map mutational effects based on changes in sequence abundance. Coupled with any of a number of selection strategies, deep mutagenesis can explore the mutational diversity available to viral glycoproteins, which mediate critical roles in cell entry and are exposed to the humoral arm of the host immune response. Mutational landscapes of viral glycoproteins for host cell attachment and membrane fusion reveal extensive epistasis and potential escape mutations to neutralizing antibodies or other therapeutics, as well as aiding in the design of optimized immunogens for eliciting broadly protective immunity. While less explored, deep mutational scans of host receptors further assist in understanding virus-host protein interactions. Critical residues on the host receptors for engaging with viral spikes are readily identified and may help with structural modeling. Furthermore, mutations may be found for engineering soluble decoy receptors as neutralizing agents that specifically bind viral targets with tight affinity and limited potential for viral escape. By untangling the complexities of how sequence contributes to viral glycoprotein and host receptor interactions, deep mutational scanning is impacting ideas and strategies at multiple levels for combatting circulating and emergent virus strains.

Mi3-GPU: MCMC-based Inverse Ising Inference on GPUs for protein covariation analysis

Inverse Ising inference is a method for inferring the coupling parameters of a Potts/Ising model based on observed site-covariation, which has found important applications in protein physics for detecting interactions between residues in protein families. We introduce Mi3-GPU ("mee-three", for MCMC Inverse Ising Inference) software for solving the inverse Ising problem for protein-sequence datasets with few analytic approximations, by parallel Markov-Chain Monte-Carlo sampling on GPUs. We also provide tools for analysis and preparation of protein-family Multiple Sequence Alignments (MSAs) to account for finite-sampling issues, which are a major source of error or bias in inverse Ising inference. Our method is "generative" in the sense that the inferred model can be used to generate synthetic MSAs whose mutational statistics (marginals) can be verified to match the dataset MSA statistics up to the limits imposed by the effects of finite sampling. Our GPU implementation enables the construction of models which reproduce the covariation patterns of the observed MSA with a precision that is not possible with more approximate methods. The main components of our method are a GPU-optimized algorithm to greatly accelerate MCMC sampling, combined with a multi-step Quasi-Newton parameter-update scheme using a "Zwanzig reweighting" technique. We demonstrate the ability of this software to produce generative models on typical protein family datasets for sequence lengths L ~ 300 with 21 residue types with tens of millions of inferred parameters in short running times.

Predominance of positive epistasis among drug resistance-associated mutations in HIV-1 protease

Drug-resistant mutations often have deleterious impacts on replication fitness, posing a fitness cost that can only be overcome by compensatory mutations. However, the role of fitness cost in the evolution of drug resistance has often been overlooked in clinical studies or in vitro selection experiments, as these observations only capture the outcome of drug selection. In this study, we systematically profile the fitness landscape of resistance-associated sites in HIV-1 protease using deep mutational scanning. We construct a mutant library covering combinations of mutations at 11 sites in HIV-1 protease, all of which are associated with resistance to protease inhibitors in clinic. Using deep sequencing, we quantify the fitness of thousands of HIV-1 protease mutants after multiple cycles of replication in human T cells. Although the majority of resistance-associated mutations have deleterious effects on viral replication, we find that epistasis among resistance-associated mutations is predominantly positive. Furthermore, our fitness data are consistent with genetic interactions inferred directly from HIV sequence data of patients. Fitness valleys formed by strong positive epistasis reduce the likelihood of reversal of drug resistance mutations. Overall, our results support the view that strong compensatory effects are involved in the emergence of clinically observed resistance mutations and provide insights to understanding fitness barriers in the evolution and reversion of drug resistance.

Structural Biology of HIV Integrase Strand Transfer Inhibitors

Integrase (IN) strand transfer inhibitors (INSTIs) are recent compounds in the antiretroviral arsenal used against HIV. INSTIs work by blocking retroviral integration; an essential step in the viral lifecycle that is catalyzed by the virally encoded IN protein within a nucleoprotein assembly called an intasome. Recent structures of lentiviral intasomes from simian immunodeficiency virus (SIV) and HIV have clarified the INSTI binding modes within the intasome active sites and helped elucidate an important mechanism of viral resistance. The structures provide an accurate depiction of interactions of intasomes and INSTIs to be leveraged for structure-based drug design. Here, we review these recent structural findings and contrast with earlier studies on prototype foamy virus intasomes. We also present and discuss examples of the latest chemical compounds that show promising inhibitory potential as INSTI candidates.

Drug Resistance Prediction Using Deep Learning Techniques on HIV-1 Sequence Data

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.