ImmuneMirror - Problem Solving Protocol - A machine learning-based integrative pipeline and web server for neoantigen prediction

Q: What problem does ImmuneMirror address in neoantigen prediction?

ImmuneMirror focuses on improving neoantigen prediction and prioritization by moving beyond binding-affinity-only assumptions, incorporating an integrative machine learning approach for more reliable results in practical settings.

Q: How was the ImmuneMirror model evaluated?

The model was trained and tested using known immunogenic neopeptides collected from 19 published studies, achieving an AUC of 0.87 on the testing data.

Q: What did ImmuneMirror reveal about MSI-H colorectal cancer patients?

It identified a subgroup of microsatellite instability-high (MSI-H) CRC patients with low neoantigen load but high tumor mutation burden, and suggested a subset with lower neoantigen load may not benefit from PD-1 blockade.

Sep 21, 2026

Uploaded by Himbo

Neoantigens originate from somatic mutations in tumors and are absent from normal tissues, making them promising immunotherapeutic targets through tumor-specific T-cell responses. ImmuneMirror is presented as a stand-alone open-source pipeline and web server that integrates a balanced random forest model for neoantigen prediction and prioritization. The model is trained and tested using known immunogenic neopeptides from 19 studies and achieves an AUC of 0.87. Application to whole-exome and RNA-seq data from 805 gastrointestinal tumors reveals an MSI-H CRC subgroup with low neoantigen load yet high mutation burden, and identifies candidates such as an HLA-A02-restricted YMCNSSCMGV-TP53G245V neopeptide in ESCC.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages12

Reading Time30 min

AccessFree to download

Introduction

Neoantigen biology and immunotherapy rationale

Existing binding-affinity tools and their limitations

ImmuneMirror pipeline and web server

Balanced random forest model for prediction and prioritization

Training and evaluation using validated neopeptides

Application to multiomics cancer cohorts

Analysis of MSI-H subgroup characteristics

Treatment relevance and immunotherapy response implications

Key findings and actionable neoantigen candidates

Briefings in Bioinformatics, 2024, 25(2), 1–12 [https://doi.org/10.1093/bib/bbae024](https://doi.org/10.1093/bib/bbae024) Problem Solving Protocol ImmuneMirror: A machine learning-based integrative pipeline and web server for neoantigen prediction Gulam Sarwar Chuwdhury†, Yunshan Guo†, Chi-Leung Chiang, Ka-On Lam, Ngar-Woon Kam, Zhonghua Liu and Wei Dai Corresponding authors: Wei Dai, Department of Clinical Oncology, Center of Cancer Medicine, School of Clinical Medicine, Li Ka Shing Faculty of Medicine, University of Hong Kong, Hong Kong (SAR), P. R. China. Tel.: (852) 39176930; Email: weidai2@hku.hku; Zhonghua Liu, Department of Biostatistics, Columbia University, 722 West 168th Street, New York, NY 10032, USA. Tel.: (001) 212 305 6604; Email: [zl2509@cumc.columbia.edu](zl2509@cumc.columbia.edu)[ ](zl2509@cumc.columbia.edu)†Gulam Sarwar Chuwdhury and Yunshan Guo contributed equally Abstract Neoantigens are derived from somatic mutations in the tumors but are absent in normal tissues. Emerging evidence suggests that neoantigens can stimulate tumor-specific T-cell-mediated antitumor immune responses, and therefore are potential immunotherapeutic targets. We developed ImmuneMirror as a stand-alone open-source pipeline and a web server incorporating a balanced random forest model for neoantigen prediction and prioritization. The prediction model was trained and tested using known immunogenic neopeptides collected from 19 published studies. The area under the curve of our trained model was 0.87 based on the testing data. We applied ImmuneMirror to the whole-exome sequencing and RNA sequencing data obtained from gastrointestinal tract cancers including 805 tumors from colorectal cancer (CRC), esophageal squamous cell carcinoma (ESCC) and hepatocellular carcinoma patients. We discovered a subgroup of microsatellite instability-high (MSI-H) CRC patients with a low neoantigen load but a high tumor mutation burden (> 10 mutations per Mbp) . Although the efficacy of PD-1 blockade has been demonstrated in advanced MSI-H patients, almost half of such patients do not respond well. Our study identified a subset of MSI-H patients who may not benefit from this treatment with lower neoantigen load for major histocompatibility complex I (P <0.0001) and II (P = 0.0008) molecules, respectively. Additionally, the neopeptide YMCNSSCMGV-TP53G245V , derived from a hotspot mutation restricted by HLA-A02, was identified as a potential actionable target in ESCC. This is so far the largest study to comprehensively evaluate neoantigen prediction models using experimentally validated neopeptides. Our results demonstrate the reliability and effectiveness of ImmuneMirror for neoantigen prediction. Keywords: neoantigen prediction; machine learning; multiomics; gastrointestinal tract cancer; immunotherapy; computational platform INTRODUCTION Immunotherapy uses the immune system to detect and fight against cancer cells. Accumulating evidence shows that the presence of neoantigens derived from somatic mutations in tumor cells elicits a potent immune response as a part of antitumor immunity through specific cytotoxic T cells [1 , 2] . Previously, various methods have been proposed for neoantigen identification, such as MHCflurry [3], NetMHCpan [4–6] and NN-Align [7], which predict the binding affinity between peptides and their corresponding major histocompatibility complex (MHC) alleles. Binding affinity is a good reference to prioritize neoantigens because MHC classes I and II help the immune system bring the bonded complex to the surface of cancerous cells for recognition by T cells. Therefore, binding to MHC molecules is a prerequisite for immunogenicity. However, the actual variant expression, human leukocyte antigen (HLA) presentation, peptide processing and transportation, as well as the ultimate T-cell response to these neoantigens, have not been considered in these existing binding affinity-based tools, therefore, these previous

FAQs

What problem does ImmuneMirror address in neoantigen prediction?

ImmuneMirror focuses on improving neoantigen prediction and prioritization by moving beyond binding-affinity-only assumptions, incorporating an integrative machine learning approach for more reliable results in practical settings.

How was the ImmuneMirror model evaluated?

The model was trained and tested using known immunogenic neopeptides collected from 19 published studies, achieving an AUC of 0.87 on the testing data.

What did ImmuneMirror reveal about MSI-H colorectal cancer patients?

It identified a subgroup of microsatellite instability-high (MSI-H) CRC patients with low neoantigen load but high tumor mutation burden, and suggested a subset with lower neoantigen load may not benefit from PD-1 blockade.

Sep 21, 2026

Uploaded by Himbo