Back

MimicNeoAI: An integrated pipeline for identifying microbial epitopes and mimicry of tumor neoepitopes

Chen, T.; Wang, W.; Zuo, X.; Zhang, Y.; Li, M.; Li, Z.; He, Y.; Zhou, Y.; Ye, F.; Zhang, B.; Jiang, Q.; Liu, H.; Zhang, L.; Fang, J.; Zhang, Y.

2025-08-09 bioinformatics
10.1101/2025.06.13.658292 bioRxiv
Show abstract

Tumor-associated microbial antigens represent promising immunotherapy targets, yet systematic identification methods remain underdeveloped. We developed MimicNeoAI, a computational pipeline integrating BiLSTM networks to identify microbial epitopes, mutation-derived neoepitopes, and their microbial mimics from sequencing data. Training on validated epitope datasets yielded 0.90 AUC with 91% accuracy on experimental validation sets. Application to colorectal cancer revealed that microbial epitopes, despite originating from a nine-fold smaller peptide pool, generated twice the immunogenic candidates (153 vs 75) compared to mutation-derived neoepitopes. These microbial epitopes exhibited exclusive tumor-specificity with no overlap in normal tissues. Single-cell TCR sequencing confirmed clonal expansion against 75% of predicted highly immunogenic epitopes, with molecular dynamics simulations demonstrating positive correlation between predicted immunogenicity and HLA-epitope-TCR binding stability. Collectively, our pipeline systematically unveils abundant, tumor-specific, and highly immunogenic microbial epitopes, providing a computational framework for developing broadly applicable cancer immunotherapies that leverage the tumor microbiome as an untapped source of therapeutic targets.

Matching journals

The top 7 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.