Predicting MHC I restricted T cell epitopes in mice with NAP-CNB, a novel online tool

Carlos Wert-Carvajal,Rubén Sánchez-García,José R Macías,Almudena Méndez Pérez,Rebeca Sanz-Pamplona,Carlos Óscar S Sorzano,Esteban Veiga,Arrate Muñoz-Barrutia,Ramon Alemany

doi:10.1038/s41598-021-89927-5

Abstract

Lack of a dedicated integrated pipeline for neoantigen discovery in mice hinders cancer immunotherapy research. Novel sequential approaches through recurrent neural networks can improve the accuracy of T-cell epitope binding affinity predictions in mice, and a simplified variant selection process can reduce operational requirements. We have developed a web server tool (NAP-CNB) for a full and automatic pipeline based on recurrent neural networks, to predict putative neoantigens from tumoral RNA sequencing reads. The developed software can estimate H-2 peptide ligands, with an AUC comparable or superior to state-of-the-art methods, directly from tumor samples. As a proof-of-concept, we used the B16 melanoma model to test the system’s predictive capabilities, and we report its putative neoantigens. NAP-CNB web server is freely available at http://biocomp.cnb.csic.es/NeoantigensApp/ with scripts and datasets accessible through the download section.

Highlights

Lack of a dedicated integrated pipeline for neoantigen discovery in mice hinders cancer immunotherapy research
Solely MuPeXI is accesible as a webserver whilst pVAC-Seq and Neoantimon have to be installed locally and require a BAM file to estimate the levels of gene expression, which underscores the importance of a comprehensive and integral pipeline as a freely accessible webservice
These tools have been trained with samples from the major histocompatibility complex (MHC) of mice or H-2. pVAC-Seq and Neoantimon include M HCflurry[16], which recently has been upgrated to include an estimation of immunogenicity through an antigen processing model using a convolutional neural network

Summary

Introduction

Lack of a dedicated integrated pipeline for neoantigen discovery in mice hinders cancer immunotherapy research. The pipelines Epi-Seq[11], pVAC-Seq3, MuPeXI9,12 and N eoantimon[13] offer modified versions for the murine model These platforms follow the canonical prediction process, based on sequencing data to estimate the gene expression and the predicted affinity with the T-cell receptor (TCR) of the mutated p eptide[10], which is a prerequisite to elicit an immune response[1]. Epi-Seq and MuPeXI use N etMHCPan[14] and its pan-specific variant, N etH2pan[15], which rely on dense neural networks for binding affinity prediction. These tools have been trained with samples from the major histocompatibility complex (MHC) of mice or H-2. Long short-term memory (LSTM) units are, at present, used for protein prediction of function and interactions[19,20]

Methods

Results

Conclusion