Understanding the genetics of viral drug resistance by integrating clinical data and mining of the scientific literature

An Goto, Raul Rodriguez‐Esteban, Sebastian H. Scharf, Garrett M. Morris

Abstract

Drug resistance caused by mutations is a public health threat for existing and emerging viral diseases. A wealth of evidence about these mutations and their clinically associated phenotypes is scattered across the literature, but a comprehensive perspective is usually lacking. This work aimed to produce a clinically relevant view for the case of Hepatitis B virus (HBV) mutations by combining a chronic HBV clinical study with a compendium of genetic mutations systematically gathered from the scientific literature. We enriched clinical mutation data by systematically mining 2,472,725 scientific articles from PubMed Central in order to gather information about the HBV mutational landscape. By performing this analysis, we were able to identify mutational hotspots for each HBV genotype (A-E) and gene (C, X, P, S), as well as the location of disulfide bonds associated with these mutations. Through a modelling study, we also identified a mutation position common in both the clinical data and the literature that is located at the binding pocket for a known anti-HBV drug, namely entecavir. The results of this novel approach show the potential of integrated analyses to assist in the development of new drugs for viral diseases that are more robust to resistance. Such analyses should be of particular interest due to the increasing importance of viral resistance in established and emerging viruses, such as for newly developed drugs against SARS-CoV-2.

Introduction

The rapid mutation of viruses like Hepatitis B (HBV) presents a significant challenge to global health, often rendering established antiviral therapies ineffective. While the scientific community generates a vast amount of data regarding these mutations, much of this information remains "scattered" across millions of publications, making it difficult for clinicians and researchers to synthesize a comprehensive view of the mutational landscape.

Clinical researchers frequently encounter novel mutations in patient samples but often lack the automated tools necessary to determine if these specific genetic changes have been previously linked to drug resistance. Conversely, mutations identified in laboratory settings may not always manifest in clinical populations. This research addresses this gap by developing an integrated, automated pipeline that synthesizes global scientific knowledge with local clinical observations to identify high-priority mutations associated with drug resistance.

Methodology

The researchers developed a multi-stage bioinformatics pipeline designed to bridge the gap between massive literature databases and real-world patient data.

Figure 1: Overview of the text mining pipeline for extracting drug-mutation co-occurrences from PMC literature.

The pipeline consists of four primary components:

  1. Text Mining: The team processed nearly 2.5 million open-access articles from PubMed Central (PMC). Utilizing rule-based text mining (regular expressions), they identified sentences where an anti-HBV drug and a specific genetic mutation co-occurred.
  2. Clinical Integration: These "literature mutations" were cross-referenced against deep-sequencing data from a Western European cohort of 186 chronic HBV patients.
  3. Normalization: To ensure accurate comparisons, a normalization step was implemented to reconcile different descriptive formats for mutations (e.g., DNA-level vs. protein-level nomenclature).
  4. Structural Modeling: Homology modeling was used to construct a 3D structure of the HBV Reverse Transcriptase. The identified mutations were then mapped onto this structure to investigate their proximity to drug-binding pockets.

Key Results and Insights

The pipeline demonstrated high efficiency, identifying 30,686 sentences containing drug-mutation pairs. From the literature corpus, 4,214 unique mutations were extracted, 316 of which (7.5%) were also present in the clinical patient data. A manual audit of 50 random sentences revealed a false-positive rate of only ~4%, indicating that rule-based mining is a robust approach for this specific domain.

Figure 3: Frequency counts of specific HBV mutations identified through the integrated analysis, categorized by protein (p.) and coding (c.) nomenclature.

The study yielded several critical insights:

  • Clinical Confirmation: 180 out of 182 patients possessed at least one mutation previously documented in the literature. This suggests that text mining can effectively validate clinical findings that might otherwise be considered noise.
  • Mutational Hotspots: The researchers identified "hotspots"—mutations appearing at significantly higher frequencies—that were consistent across HBV genotypes A through E. These hotspots represent primary targets for future drug development.
  • Structural Vulnerabilities: Mapping mutations to 3D models revealed clusters around the binding pocket of Entecavir. Additionally, mutations in cysteine residues (C124 and C149) were identified as likely to destabilize viral protein structures by disrupting disulfide bonds.

Discussion and Future Directions

This integrated approach provides a powerful framework for automated hypothesis generation. By identifying mutations that co-occur with drugs in the literature, researchers can generate a prioritized list of candidates for laboratory validation.

However, the authors note that simple co-occurrence serves as a proxy for resistance rather than definitive proof. Future work will focus on expanding the corpus to include paywalled journals and utilizing advanced machine learning-based Natural Language Processing (NLP) to capture more nuanced descriptions of mutational changes. Furthermore, developing techniques to extract data from tables and figures will be essential, as significant genetic information is often stored in non-textual formats.

The framework's adaptability suggests it could be rapidly deployed for other viral threats, including SARS-CoV-2, to monitor the emergence of resistance to new antivirals.

Summary updated

Abstract

Drug resistance caused by mutations is a public health threat for existing and emerging viral diseases. A wealth of evidence about these mutations and their clinically associated phenotypes is scattered across the literature, but a comprehensive perspective is usually lacking. This work aimed to produce a clinically relevant view for the case of Hepatitis B virus (HBV) mutations by combining a chronic HBV clinical study with a compendium of genetic mutations systematically gathered from the scientific literature. We enriched clinical mutation data by systematically mining 2,472,725 scientific articles from PubMed Central in order to gather information about the HBV mutational landscape. By performing this analysis, we were able to identify mutational hotspots for each HBV genotype (A-E) and gene (C, X, P, S), as well as the location of disulfide bonds associated with these mutations. Through a modelling study, we also identified a mutation position common in both the clinical data and the literature that is located at the binding pocket for a known anti-HBV drug, namely entecavir. The results of this novel approach show the potential of integrated analyses to assist in the development of new drugs for viral diseases that are more robust to resistance. Such analyses should be of particular interest due to the increasing importance of viral resistance in established and emerging viruses, such as for newly developed drugs against SARS-CoV-2.
BibTeX
@inproceedings{goto2022understanding,
  author = {An Goto and Raul Rodriguez‐Esteban and Sebastian H. Scharf and Garrett M. Morris},
  title = {{Understanding the genetics of viral drug resistance by integrating clinical data and mining of the scientific literature}},
  booktitle = {Scientific},
  year = {2022},
  doi = {10.1038/s41598-022-17746-3},
}
Figure