Assessing off-target toxicity of cellular immunotherapies - CIMT 2021 poster
Assessing off-target toxicity of cellular immunotherapies
We have the pleasure to present our poster at CIMT 2021. Please read the abstract below and at the end, you can see our poster and video explanations of our research.
Despite their great clinical potential, cell-based cancer immunotherapies are characterized by a high risk of generating off-target immunotoxicity against healthy tissues, which can have serious health consequences for patients undergoing such therapies. Unfortunately, identifying the epitopes responsible for inducing these off-target toxicities using experimental methods is costly, time-consuming, and limited to a restricted subspace of sequences, since probing the entire space experimentally is practically unfeasible. Here, we present ArdImmune Tox – a computational tool capable of evaluating potential off-target toxicities by leveraging recent advances in computational immunology and Artificial Intelligence (AI).
Given a target peptide and its associated HLA type, a large collection of putative off-target epitopes (OTEs) having up to a predefined number of amino acid differences with respect to the original peptide, is generated computationally. Putative OTEs are then filtered according to their predicted probability of HLA presentation. Then, the probability that each amino acid interacts with the TCR is evaluated for each OTE and the TCR-facing residues are compared with the corresponding ones in the target peptide, based on selected physico-chemical descriptors. Next, the genomic position of each OTE is annotated and used to extract frequent variants from a dataset encompassing approximately 76,000 genomes. Performing this step, several alternative OTE variants are obtained. The expression patterns (both related to mRNA and protein) of OTEs are reported, as highly expressed off-targets are more likely to induce toxicity.
We evaluated the predictive performance of ArdImmune Tox on three OTEs obtained from the literature, all having experimentally confirmed off-target toxicity. In all cases, our method correctly identified the OTEs ranking them among the two highest scoring potential off-target peptides. We compared our results with those produced by other methods in which the OTEs were positioned much lower in the rank list or were missing altogether. d) Statement of the conclusions Our results suggest that ArdImmune Tox represents an effective computational approach for the identification of peptide- based off-target toxicity and cross-reactivity of T lymphocytes, with potential applications in the development of immunotherapies and in the study of molecular mimicry. We believe ArdImmune Tox may help improve the safety profile of epitope-based cancer immunotherapies.