ImmuneMirror

ImmuneMirror incorporates a balanced random forest model for neoantigen prediction and prioritization to identify immunotherapeutic targets from tumor sequencing data.

Key Features:

• Balanced Random Forest Model: The prediction model was trained and tested using known immunogenic neopeptides collected from 19 published studies, achieving an area under the curve of 0.87 on testing data.
• Neoantigen Prediction and Prioritization: Predicts neoantigens derived from somatic mutations present in tumors but absent in normal tissues.
• Whole-Exome Sequencing and RNA Sequencing Integration: Processes WES and RNA-seq data for neoantigen identification.
• Major Histocompatibility Complex I and II Analysis: Evaluates neoantigen load for both MHC class I and MHC class II molecules.

Scientific Applications:

• Gastrointestinal Tract Cancer Analysis: Applied to 805 tumors from colorectal cancer, esophageal squamous cell carcinoma, and hepatocellular carcinoma patients.
• Microsatellite Instability-High Subgroup Identification: Discovered a subgroup of MSI-H colorectal cancer patients with low neoantigen load but high tumor mutation burden, who may not respond well to PD-1 blockade.
• Actionable Target Discovery: Identified the neopeptide YMCNSSCMGV-TP53G245V, derived from a hotspot mutation restricted by HLA-A02, as a potential actionable target in esophageal squamous cell carcinoma.
• Model Validation: The largest study to comprehensively evaluate neoantigen prediction models using experimentally validated neopeptides.

Methodology:

The balanced random forest model was trained and tested using known immunogenic neopeptides collected from 19 published studies.