Structure-guided cancer immunotherapy design with HLA-Arena and CrossDome

Project Summary
The broader use of T-cell-based therapies is still hindered by challenges related to the identification of peptide-
targets that are both immunogenic (capable of activating T-cells) and safe (do not trigger on-target/off-tumor or
off-target toxicities). This is in part due to persistent dependency on biased sequence-based methods, despite
recent breakthroughs in structural modeling and machine learning that could be leveraged to support new workflows
for the identification of tumor-associated antigens (TAAs). To address this issue, and foster the design of better
T-cell-based immunotherapies, we propose a new computational environment (HLA-arena 2.0) that will integrate
existing ITCR resources, with new bioinformatics methods for structural modeling and analysis of key cellular
immunity receptors; namely T-cell receptors (TCRs) and Human Leukocyte Antigen (HLA) receptors. Our working
hypothesis is that the combination of multi-omics data with large-scale structure-based analysis can overcome
most of the limitations of existing pipelines for TAA discovery, therefore enabling the design of better and safer
T-cell-based immunotherapies. To test this hypothesis, we will implement a new workflow for structure-guided
TAA discovery, integrating HLA-Arena with pVACtools (ITCR-funded package for sequence-based neoantigen
discovery) and CrossDome (an R package for off-target toxicity prediction). In collaboration with researchers
from MD Anderson Cancer Center, the PI will develop and test workflows to address existing needs in T-cell-
based immunotherapy. We will focus on two different cancer types, that represent different challenges for cancer
immunotherapy. In collaboration with Dr. Lizée, we will benchmark our structure-guided TAA discovery workflow
using immunopeptidomics data on melanoma. We will also run off-target toxicity predictions to identify the safest
among 10 potentially therapeutic T-cell clones targeting two melanoma-derived TAAs from SLC45A2. Melanoma is
a type of solid tumor for which greater success has been observed with immunotherapy treatments. On the other
hand, acute myeloid leukemia (AML) is a type of blood cancer in which severe reactions to immunotherapy have
been observed. In this context, we will work with Dr. Abbas to examine transcriptomic datasets (bulk and single-cell
data) from AML patients, aiming at uncovering TAAs and TCRs that are associated with effective immune response
to AML. Finally, we will use CrossDome and existing data on known TAAs to develop The Cancer off-target Toxicity
Atlas (TCTA). For each known TAA, this new database will contain a list of potential off-targets that should be tested
when targeting these TAAs with immunotherapies. Predicted off-targets will be annotated with additional data (e.g.,
tissue expression, HLA-binding, immunogenicity, etc). All methods will be made available to the community through
user-friendly workflows, facilitating the design of better and safer T-cell-based immunotherapies for numerous types
of cancer. The proposed methods will be deeply integrated into the ITCR network, creating many opportunities for
future collaborations. In addition, the long-term goals of the proposed research are well aligned with NCI’s mission
to achieve more effective and less toxic cancer treatments, therefore helping people live longer and healthier lives.

Grant Number: 1R21CA289333-01
NIH Institute/Center: NIH
Principal Investigator: Dinler Antunes