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 mutational 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 shows 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 Challenge of Viral Resistance
Viral drug resistance is a moving target. While clinical studies frequently identify new genetic mutations in patients, determining whether these mutations actually confer resistance is a monumental task. The evidence is often "scattered across the literature," buried in millions of papers that are impossible for a human researcher to synthesize manually. For Hepatitis B—a virus affecting 292 million people—this gap is critical. Chronic HBV can lead to cirrhosis and liver cancer, and while treatments like nucleotide analogues exist, the virus’s high mutation rate allows it to bypass these drugs. This research addresses the need for a comprehensive perspective by bridging the gap between raw clinical observations and the vast, unstructured wealth of existing scientific knowledge.
Methodology: A Hybrid Data Integration Approach
The researchers developed a multi-step integrated analysis pipeline to synthesize clinical and literature-based data. The core of this approach was a rule-based text-mining system designed to extract meaningful associations from a massive corpus of scientific literature.
- Text Mining: Using Python-based tools, the team processed 2,472,725 XML files from PubMed Central (PMC). They employed a rule-based approach to identify sentences where an approved HBV drug and a mutation co-occurred.
- Normalization: Because scientific papers use inconsistent naming conventions (e.g., DNA vs. amino acid variants), a secondary pipeline was built to normalize these into a standard format (c. and p. forms) for direct comparison with clinical data.
- Clinical Integration: The literature-mined data was compared against ultra-deep sequencing results from a Western European cohort of 186 HBV patients.
- Structural Modeling: To understand the functional impact of identified mutations, the team performed homology modeling to map mutations onto the 3D structure of the HBV Reverse Transcriptase (RT) protein, specifically focusing on the binding pocket of the drug entecavir.
Key Results: Mapping the Mutational Landscape
The pipeline identified 30,686 publications containing co-occurrences of HBV drugs and mutations, resulting in 4,214 unique mutations. Approximately 7.5% of the mutations found in the literature were also present in the clinical cohort, providing immediate context to clinical observations.
In the clinical study, the most frequent amino acid mutation, p.Phe220Leu, appeared in 52 patients, while the most frequent nucleotide mutation, c.389A>T, appeared in 78 patients. The study successfully mapped these hotspots across all four HBV genes (C, X, P, S) and five genotypes (A-E), creating a visual landscape of resistance.
Discussion: Insights into Resistance Mechanisms
- The "Bridge" Utility: The study demonstrates that text mining can validate clinical findings in real-time. 180 out of 182 patients had mutations that had already been documented in the literature.
- Genotype-Common Hotspots: The analysis identified "genotype-common" mutation hotspots—positions like K293 in the polymerase or S213 in the reverse transcriptase—that appear across multiple HBV genotypes. These may represent universal adaptation strategies for the virus.
- Structural Vulnerabilities: Many common mutations occur at disulfide bond sites (C124 and C149), which are vital for protein stability. These mutations likely destabilize viral proteins, potentially altering drug interactions.
- The Entecavir Pocket: By mapping clinical/literature overlaps onto a 3D model, the researchers identified specific mutations (I169, M204, N238) located within 12 Å of the entecavir binding site, providing a structural rationale for drug interference.
Limitations and Future Work
While promising, the 7.5% overlap suggests that much of the mutational landscape remains undocumented or is hidden in non-PMC repositories. Furthermore, the current co-occurrence rule can flag non-resistance-associated mutations. Future iterations should utilize advanced NLP to better understand the context of mentions. Despite these limitations, this framework is highly applicable to other viruses, including SARS-CoV-2, where understanding resistance is of paramount importance.
Summary updated