MimicNeoAI: An integrated pipeline for identifying microbial mimicry antigens and tumor neoantigen partners

Read the full article See related articles

Listed in

This article is not in any list yet, why not save it to one of your lists.
Log in to save this article

Abstract

Abstract Recent studies underscore the significant role of the tumor microbiome, defined as the microbial community within the tumor microenvironment, in influencing cancer development, progression, and treatment response. Notably, epitopes derived from these tumor-resident microbial species can mimic tumor neoantigens through cross-reactive mechanisms, a phenomenon termed molecular mimicry. Unlike neoantigens, which are highly patient-specific, microbial antigens exhibit lower personalization, as shared microbial species are frequently enriched across various patients and cancer types. This conservation enables the identification of recurrent microbial antigen mimics that are common to multiple cancers, thereby providing potential targets for broad-spectrum immunotherapies. Despite the growing recognition of microbial antigens in cancer immunity, the field currently lacks dedicated computational tools for systematically characterizing tumor-associated microbial antigens and their potential mimicry candidates. To address this gap, we present MimicNeoAI, a computational pipeline that integrates sequencing data to concurrently identify microbial antigens and tumor neoantigens. Additionally, it predicts the immunogenic potential of peptide- Human Leukocyte Antigen (HLA) complexes using a Bidirectional Long Short-Term Memory (BiLSTM) network model, which is trained on immunogenic epitope datasets from both hosts and microbes. Furthermore, this pipeline assesses microbial mimicry antigens and tumor neoantigen pairs through sequence similarity, allowing for the identification of candidates that may elicit cross-reactive immune responses and enhance immunotherapy. In cross-validation, MimicNeoAI achieved an average AUC of 0.90, successfully predicting 91% of microbial antigen-positive data. When applied to colorectal cancer (CRC), it identified five high-confidence microbial mimicry antigens associated with increased T cell clonal expansion, as evidenced by single-cell T Cell Receptor (TCR) sequencing, highlighting their biological relevance and potential as targets for immunotherapy.

Article activity feed