DeepImmuno: Deep learning-empowered prediction and generation of immunogenic peptides for T cell immunity

Abstract

T-cells play an essential role in the adaptive immune system by seeking out, binding and destroying foreign antigens presented on the cell surface of diseased cells. An improved understanding of T-cell immunity will greatly aid in the development of new cancer immunotherapies and vaccines for life threatening pathogens. Central to the design of such targeted therapies are computational methods to predict non-native epitopes to elicit a T cell response, however, we currently lack accurate immunogenicity inference methods. Another challenge is the ability to accurately simulate immunogenic peptides for specific human leukocyte antigen (HLA) alleles, for both synthetic biological applications and to augment real training datasets. Here, we proposed a beta-binomial distribution approach to derive epitope immunogenic potential from sequence alone. We conducted systematic benchmarking of five traditional machine learning (ElasticNet, KNN, SVM, Random Forest, AdaBoost) and three deep learning models (CNN, ResNet, GNN) using three independent prior validated immunogenic peptide collections (dengue virus, cancer neoantigen and SARS-Cov-2). We chose the CNN model as the best prediction model based on its adaptivity for small and large datasets, and performance relative to existing methods. In addition to outperforming two highly used immunogenicity prediction algorithms, DeepHLApan and IEDB, DeepImmuno-CNN further correctly predicts which residues are most important for T cell antigen recognition. Our independent generative adversarial network (GAN) approach, DeepImmuno-GAN, was further able to accurately simulate immunogenic peptides with physiochemical properties and immunogenicity predictions similar to that of real antigens. We provide DeepImmuno-CNN as source code and an easy-to-use web interface.Data AvailabilityDeepImmuno Python3 code is available at https://github.com/frankligy/DeepImmuno. The DeepImmuno web portal is available from https://deepimmuno.herokuapp.com. The data in this article is available in GitHub and supplementary materials.

Evaluating performance of existing computational models in predicting CD8+ T cell pathogenic epitopes and cancer neoantigens

“…To perform an unbiased evaluation of existing models in predicting immunogenic peptides, we identified 7 publicly available models (see Methods for criteria details), each of which aim to predict whether an MHC-presented peptide may invoke a T cell response (i.e., whether a peptide is immunogenic). These models are named as the IEDB model 16 , NetTepi 15 , iPred 14 , Repitope 27 , PRIME 22 , DeepImmuno 17 and Gao 28 (Table 1). Additionally, given the observation by Paul et al 29 that netMHCpan 4.0 most accurately identified immunogenic vaccinia-virus T cell epitopes in mice, we included both eluted ligand (labelled netMHCpan_EL) and binding affinity (labelled netMHCpan_BA) outputs from netMHCpan 4.0 to assess their accuracy in a human setting.…”

“…Click or tap here to enter text.Click or tap here to enter text.Despite these challenges, over the past decade several models have been presented to predict immunogenic peptides, leveraging different correlates of immunogenicity with varying levels of success. As we recently detailed 8 , a number of these studies have utilised sequence-based characteristics including amino acid features [13][14][15][16][17] , similarity to viral peptides 18 , sequence dissimilarity to self 3,19 , association between peptide immunogenicity and their biophysical properties such as their structural and energy features 20 , as well as TCR recognition features 13,21 . Recently, Wells et al 3 comprehensively investigated a collection of parameters associated with neoantigen immunogenicity, grouped into 1) presentation features e.g., binding affinity and stability, hydrophobicity and tumor abundance and 2) recognition features e.g., agretopicity (the ratio of binding affinity between a mutated peptide and its wild-type counterpart), and foreignness (similarity of peptide of interest to previously characterised viral epitopes).…”

Enhancing Our Understanding of Plant Cell-to-Cell Interactions Using Single-Cell Omics

“…Increasing the complexity of these interactions, the neighboring plant cells to an event of infection will trigger a non-cell-autonomous immunity response to minimize new events of infection by the same microbe. This immunity is activated upon communication between plant cells (Yan et al, 2019;Aung et al, 2020;Li et al, 2020;Zeng et al, 2020). Considering that different events of interaction and infection co-occur in a complex organ, the molecular characterization of the cell-autonomous and non-cell-autonomous responses of plant cells to a pathogenic infection remains challenging when conducted at the level of complex tissues and organs.…”

Spread and Dynamics of COVID-19 in the Kingdom of Saudi Arabia and Four Other Countries in the Early Phase: A Study for the Development of the Advanced Health System

“…In silico provides a unique generative adversarial network (GAN) based approach for synthesizing new drugs for speci c diseases including viral diseases (COVID-19) (Ibrahim et al, 2020;Li et al, 2021;. Their network allows for de novo small molecule design, which optimizes synthetic feasibility, novelty, and biological activity.…”

“…However, few examples of generative drug design have achieved experimental validation in vitro or in vivo. The in silico approaches recently published 10 representative structures of protease inhibitors for their potential development against COVID19 (Keretsu et al, 2020;Li et al, 2021).…”