A Computational Strategy for the Rapid Identification and Ranking of Patient-Specific T cell Receptors Bound to Neoantigens.

T Cell Receptor (TCR) recognition of a peptide-major histocompatibility complex (pMHC) is crucial for adaptive immune response. The identification of therapeutically relevant TCR-pMHC protein pairs is a bottleneck in the implementation of TCR-based immunotherapies. The ability to computationally design TCRs to target a specific pMHC requires automated integration of next-generation sequencing, protein-protein structure prediction, molecular dynamics, and TCR ranking. We present a pipeline to evaluate patient-specific, sequence-based TCRs to a target pMHC. Using the three most frequently expressed TCRs from 16 colorectal cancer patients, we predict the protein-protein structure of the TCRs to the target CEA peptide-MHC using Modeller and ColabFold. TCR-pMHC structures are compared using automated equilibration and successive analysis. ColabFold generated configurations require a ∼2.5X reduction in equilibration time of TCR-pMHC structures compared to Modeller. The structural differences between Modeller and ColabFold are demonstrated by root mean square deviation (∼0.20 nm) between clusters of equilibrated configurations, which impact the number of hydrogen bonds and Lennard-Jones contacts between the TCR and pMHC. We identify TCR ranking criteria that may prioritize TCRs for evaluation of in vitro immunogenicity and validate this ranking by comparing to state-of-the-art machine learning based methods trained to predict the probability of TCR-pMHC binding. This article is protected by copyright. All rights reserved.

MolPROP: Molecular Property prediction with multimodal language and graph fusion.

Pretrained deep learning models self-supervised on large datasets of language, image, and graph representations are often fine-tuned on downstream tasks and have demonstrated remarkable adaptability in a variety of applications including chatbots, autonomous driving, and protein folding. Additional research aims to improve performance on downstream tasks by fusing high dimensional data representations across multiple modalities. In this work, we explore a novel fusion of a pretrained language model, ChemBERTa-2, with graph neural networks for the task of molecular property prediction. We benchmark the MolPROP suite of models on seven scaffold split MoleculeNet datasets and compare with state-of-the-art architectures. We find that (1) multimodal property prediction for small molecules can match or significantly outperform modern architectures on hydration free energy (FreeSolv), experimental water solubility (ESOL), lipophilicity (Lipo), and clinical toxicity tasks (ClinTox), (2) the MolPROP multimodal fusion is predominantly beneficial on regression tasks, (3) the ChemBERTa-2 masked language model pretraining task (MLM) outperformed multitask regression pretraining task (MTR) when fused with graph neural networks for multimodal property prediction, and (4) despite improvements from multimodal fusion on regression tasks MolPROP significantly underperforms on some classification tasks. MolPROP has been made available at https://github.com/merck/MolPROP . SCIENTIFIC CONTRIBUTION: This work explores a novel multimodal fusion of learned language and graph representations of small molecules for the supervised task of molecular property prediction. The MolPROP suite of models demonstrates that language and graph fusion can significantly outperform modern architectures on several regression prediction tasks and also provides the opportunity to explore alternative fusion strategies on classification tasks for multimodal molecular property prediction.

AbLEF: antibody language ensemble fusion for thermodynamically empowered property predictions.

MOTIVATION: Pre-trained protein language and/or structural models are often fine-tuned on drug development properties (i.e. developability properties) to accelerate drug discovery initiatives. However, these models generally rely on a single structural conformation and/or a single sequence as a molecular representation. We present a physics-based model, whereby 3D conformational ensemble representations are fused by a transformer-based architecture and concatenated to a language representation to predict antibody protein properties. Antibody language ensemble fusion enables the direct infusion of thermodynamic information into latent space and this enhances property prediction by explicitly infusing dynamic molecular behavior that occurs during experimental measurement. RESULTS: We showcase the antibody language ensemble fusion model on two developability properties: hydrophobic interaction chromatography retention time and temperature of aggregation (Tagg). We find that (i) 3D conformational ensembles that are generated from molecular simulation can further improve antibody property prediction for small datasets, (ii) the performance benefit from 3D conformational ensembles matches shallow machine learning methods in the small data regime, and (iii) fine-tuned large protein language models can match smaller antibody-specific language models at predicting antibody properties. AVAILABILITY AND IMPLEMENTATION: AbLEF codebase is available at https://github.com/merck/AbLEF.

A microfluidic strategy to capture antigen-specific high affinity B cells.

Assessing B cell affinity to pathogen-specific antigens prior to or following exposure could facilitate the assessment of immune status. Current standard tools to assess antigen-specific B cell responses focus on equilibrium binding of the secreted antibody in serum. These methods are costly, time-consuming, and assess antibody affinity under zero-force. Recent findings indicate that force may influence BCR-antigen binding interactions and thus immune status. Here, we designed a simple laminar flow microfluidic chamber in which the antigen (hemagglutinin of influenza A) is bound to the chamber surface to assess antigen-specific BCR binding affinity of five hemagglutinin-specific hybridomas under 65- to 650-pN force range. Our results demonstrate that both increasing shear force and bound lifetime can be used to enrich antigen-specific high affinity B cells. The affinity of the membrane-bound BCR in the flow chamber correlates well with the affinity of the matched antibodies measured in solution. These findings demonstrate that a microfluidic strategy can rapidly assess BCR-antigen binding properties and identify antigen-specific high affinity B cells. This strategy has the potential to both assess functional immune status from peripheral B cells and be a cost-effective way of identifying individual B cells as antibody sources for a range of clinical applications.

A dynamic biomimetic model of the membrane-bound CD4-CD3-TCR complex during pMHC disengagement.

The coordinated (dis)engagement of the membrane-bound T cell receptor (TCR)-CD3-CD4 complex from the peptide-major histocompatibility complex (pMHC) is fundamental to TCR signal transduction and T cell effector function. As such, an atomic-scale understanding would not only enhance our basic understanding of the adaptive immune response but would also accelerate the rational design of TCRs for immunotherapy. In this study, we explore the impact of the CD4 coreceptor on the TCR-pMHC (dis)engagement by constructing a molecular-level biomimetic model of the CD3-TCR-pMHC and CD4-CD3-TCR-pMHC complexes within a lipid bilayer. After allowing the system complexes to equilibrate (engage), we use steered molecular dynamics to dissociate (disengage) the pMHC. We find that 1) the CD4 confines the pMHC closer to the T cell by 1.8 nm at equilibrium; 2) CD4 confinement shifts the TCR along the MHC binding groove engaging a different set of amino acids and enhancing the TCR-pMHC bond lifetime; 3) the CD4 translocates under load increasing the interaction strength between the CD4-pMHC, CD4-TCR, and CD4-CD3; and 4) upon dissociation, the CD3-TCR complex undergoes structural oscillation and increased energetic fluctuation between the CD3-TCR and CD3-lipids. These atomic-level simulations provide mechanistic insight on how the CD4 coreceptor impacts TCR-pMHC (dis)engagement. More specifically, our results provide further support (enhanced bond lifetime) for a force-dependent kinetic proofreading model and identify an alternate set of amino acids in the TCR that dominate the TCR-pMHC interaction and could thus impact the design of TCRs for immunotherapy.

Receptor-ligand non-equilibrium kinetics (RLNEK) 1.0: An integrated Trackmate laminar flow chamber analysis.

Although parallel plate flow chamber assays are widely performed, extraction of kinetic parameters is limited to specialized labs with mathematical expertise and customized video-microscopy tracking tools. The recent development of Trackmate has increased researcher accessibility to tracking particles in video-microscopy experiments; however, there is a lack of tools that analyze this tracking information. We report a software tool, compatible with Trackmate, that extracts Receptor Ligand Non-Equilibrium Kinetic (RLNEK) parameters from video-microscopy data. This software should be of particular interest to the community of researchers and scientists interrogating the target-specific binding and release of immune cells.

A computational algorithm to assess the physiochemical determinants of T cell receptor dissociation kinetics.

The rational design of T Cell Receptors (TCRs) for immunotherapy has stagnated due to a limited understanding of the dynamic physiochemical features of the TCR that elicit an immunogenic response. The physiochemical features of the TCR-peptide major histocompatibility complex (pMHC) bond dictate bond lifetime which, in turn, correlates with immunogenicity. Here, we: i) characterize the force-dependent dissociation kinetics of the bond between a TCR and a set of pMHC ligands using Steered Molecular Dynamics (SMD); and ii) implement a machine learning algorithm to identify which physiochemical features of the TCR govern dissociation kinetics. Our results demonstrate that the total number of hydrogen bonds between the CDR2ß-MHC⍺(ß), CDR1α-Peptide, and CDR3ß-Peptide are critical features that determine bond lifetime.

Using molecular dynamics simulations to interrogate T cell receptor non-equilibrium kinetics.

An atomic-scale mechanism of T Cell Receptor (TCR) mechanosensing of peptides in the binding groove of the peptide-major histocompatibility complex (pMHC) may inform the design of novel TCRs for immunotherapies. Using steered molecular dynamics simulations, our study demonstrates that mutations to peptides in the binding groove of the pMHC - which are known to discretely alter the T cell response to an antigen - alter the MHC conformation at equilibrium. This subsequently impacts the overall strength (duration and length) of the TCR-pMHC bond under constant load. Moreover, physiochemical features of the TCR-pMHC dynamic bond strength, such as hydrogen bonds and Lennard-Jones contacts, correlate with the immunogenic response elicited by the specific peptide in the MHC groove. Thus, formation of transient TCR-pMHC bonds is characteristic of immunogenic peptides, and steered molecular dynamics simulations can be used in the overall design strategy of TCRs for immunotherapies.

Organ-on-a-chip model of vascularized human bone marrow niches.

Bone marrow niches (endosteal and perivascular) play important roles in both normal bone marrow function and pathological processes such as cancer cell dormancy. Unraveling the mechanisms underlying these events in humans has been severely limited by models that cannot dissect dynamic events at the niche level. Utilizing microfluidic and stem cell technologies, we present a 3D in vitro model of human bone marrow that contains both the perivascular and endosteal niches, complete with dynamic, perfusable vascular networks. We demonstrate that our model can replicate in vivo bone marrow function, including maintenance and differentiation of CD34+ hematopoietic stem/progenitor cells, egress of neutrophils (CD66b+), and niche-specific responses to doxorubicin and granulocyte-colony stimulating factor. Our platform provides opportunities to accelerate current understanding of human bone marrow function and drug response with high spatial and temporal resolution.