Helicopredict

Helicobacter pylori (H. pylori) infection is one of the most prevalent bacterial infections worldwide. Chronic infection leads to chronic active gastritis and can result in the development of other complications such as ulcers or gastric cancer. Indeed, approximately 90% of all gastric cancers are associated with H. pylori. Failure of standard eradication therapies is rising dramatically due to the increased development of resistant bacterial strains. Since two antibiotics are needed for successful eradication, the usage of only one antibiotic in other indications such as respiratory diseases will render the (mostly yet undetected) H. pylori strain in these patients resistant. Nowadays, it is estimated that already 10-20% of H. pylori strains are multiresistant. However, culture-based resistance testing, which is currently only recommended after unsuccessful second-line therapy, is a lengthy process; in vitro growth of H. pylori takes 5-7 days after isolation from gastric tissue and further resistance tests last between 3 to 5 days. Taking this into account, a rapid method to determine whether an isolated strain will be prone to antibiotic resistance would be a tremendous help to choose the appropriate therapeutic regimen.To address this challenge, we plan to develop an algorithm for prediction of antibiotic resistance, which will primarily be based on H. pylori genome sequencing data that can be obtained quickly. The algorithm will be made publicly available to physicians and serve as a way to select the optimal therapy. This approach will help to optimize therapeutic efficacy and counteract further resistance development.

Within the funding period, the Helicopredict network partners created a scientific foundation that had never existed in this form before: a unique collection of more than 500 clinical H. pylori isolates, each thoroughly tested for resistance to the three most clinically relevant antibiotics—clarithromycin, levofloxacin, and metronidazole. Every isolate was fully genome-sequenced using advanced technologies such as Illumina MiSeq and nanopore sequencing, and complemented by detailed microbiological analyses including E-tests and PCR-based methods. Based on this high-quality dataset, the team developed next-generation AI models that provide accurate and clinically relevant resistance predictions. The most striking progress was achieved for metronidazole resistance, long considered difficult to predict. Here, the models not only confirmed known mechanisms but also revealed previously undescribed mutations, significantly deepening our understanding of bacterial adaptation and opening new avenues for research and therapy design.

In parallel, the partners built a modern digital infrastructure that integrates clinical, microbiological, and genomic data across institutions and makes them usable for both research and clinical applications. Working closely with an industry partner, they developed a web-based platform that enables a future shift towards genome-guided therapy: instead of empirical antibiotic selection, a single genome test could soon be sufficient to recommend the most effective treatment for each patient, reduce unnecessary antibiotic use, and improve eradication success rates. The system has already been engineered to incorporate additional data layers, including microbiome datasets from TUM. Despite pandemic-related challenges, the network partners achieved all project milestones and established a foundation that strengthens Munich’s more than 25-year international leadership in H. pylori research—while setting the stage for long-term global impact on diagnostics and treatment strategies.