An official website of the United States government
Abstract The ability to accurately identify peptide ligands for a given major histocompatibility complex class I (MHC-I) molecule has immense value for targeted anticancer therapeutics. However, the highly polymorphic nature of the MHC-I protein makes universal prediction of peptide ligands challenging due to lack of experimental data describing most MHC-I variants. To address this challenge, we have developed a deep convolutional neural network, HLA-Inception, capable of predicting MHC-I peptide binding motifs using electrostatic properties of the MHC-I binding pocket. By approaching this immunological issue using molecular biophysics, we measure the impact of sidechain arrangement and topology on peptide binding, feature not captured by sequence-based MHC-I prediction methods. Through a combination of molecular modeling and simulation, 5821 MHC-I alleles were modeled, providing extensive coverage across human populations. Predicted peptide binding motifs fell into distinct clusters, each defined with different degrees of submotif heterogeneity. Peptide binding scores generated by HLA-Inception are strongly correlated with quantitative MHC-I binding data, indicating predicted peptides can be ranked, both within and between alleles. HLA-inception also showed high precision when predicting naturally presented peptides and can be used for rapid proteome-scale MHC-I peptide binding predictions. Finally, we show that the binding pocket diversity measured by HLA inception predicts response to checkpoint blockade. Citation Format: Eric A. Wilson, John Kevin Cava, Diego Chowell, Abhishek Singharoy, Karen S. Anderson. Protein structure-based modeling to improve MHC class I epitope predictions. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 5376.
more »
« less
Evaluating NetMHCpan performance on non-European HLA alleles not present in training data
Bias in neural network model training datasets has been observed to decrease prediction accuracy for groups underrepresented in training data. Thus, investigating the composition of training datasets used in machine learning models with healthcare applications is vital to ensure equity. Two such machine learning models are NetMHCpan-4.1 and NetMHCIIpan-4.0, used to predict antigen binding scores to major histocompatibility complex class I and II molecules, respectively. As antigen presentation is a critical step in mounting the adaptive immune response, previous work has used these or similar predictions models in a broad array of applications, from explaining asymptomatic viral infection to cancer neoantigen prediction. However, these models have also been shown to be biased toward hydrophobic peptides, suggesting the network could also contain other sources of bias. Here, we report the composition of the networks’ training datasets are heavily biased toward European Caucasian individuals and against Asian and Pacific Islander individuals. We test the ability of NetMHCpan-4.1 and NetMHCpan-4.0 to distinguish true binders from randomly generated peptides on alleles not included in the training datasets. Unexpectedly, we fail to find evidence that the disparities in training data lead to a meaningful difference in prediction quality for alleles not present in the training data. We attempt to explain this result by mapping the HLA sequence space to determine the sequence diversity of the training dataset. Furthermore, we link the residues which have the greatest impact on NetMHCpan predictions to structural features for three alleles (HLA-A*34:01, HLA-C*04:03, HLA-DRB1*12:02).
more »
« less
- Award ID(s):
- 2036064
- PAR ID:
- 10489972
- Publisher / Repository:
- Frontiers in Immunology
- Date Published:
- Journal Name:
- Frontiers in Immunology
- Volume:
- 14
- ISSN:
- 1664-3224
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
George Bebis, Terry Gaasterland (Ed.)Major Histocompability Complex (MHC) Class I molecules provide a pathway for cells to present endogenous peptides to the immune system, allowing it to distinguish healthy cells from those infected by pathogens. Software tools based on neural networks such as NetMHC and NetMHCpan predict whether peptides will bind to variants of MHC molecules. These tools are trained with experimental data, consisting of the amino acid sequence of peptides and their observed binding strength. Such tools generally do not explicitly consider hydrophobicity, a significant biochemical factor relevant to peptide binding. It was observed that these tools predict that some highly hydrophobic peptides will be strong binders, which biochemical factors suggest is incorrect. This paper investigates the correlation of the hydrophobicity of 9-mer peptides with their predicted binding strength to the MHC variant HLA-A*0201 for these software tools. Two studies were performed, one using the data that the neural networks were trained on and the other using a sample of the human proteome. A significant bias within NetMHC-4.0 towards predicting highly hydrophobic peptides as strong binders was observed in both studies. This suggests that hydrophobicity should be included in the training data of the neural networks. Retraining the neural networks with such biochemical annotations of hydrophobicity could increase the accuracy of their predictions, increasing their impact in applications such as vaccine design and neoantigen identification.more » « less
-
Reliable prediction of T cell specificity against antigenic signatures is a formidable task, complicated by the immense diversity of T cell receptor and antigen sequence space and the resulting limited availability of training sets for inferential models. Recent modeling efforts have demonstrated the advantage of incorporating structural information to overcome the need for extensive training sequence data, yet disentangling the heterogeneous TCR-antigen interface to accurately predict MHC-allele-restricted TCR-peptide interactions has remained challenging. Here, we present RACER-m, a coarse-grained structural model leveraging key biophysical information from the diversity of publicly available TCR-antigen crystal structures. Explicit inclusion of structural content substantially reduces the required number of training examples and maintains reliable predictions of TCR-recognition specificity and sensitivity across diverse biological contexts. Our model capably identifies biophysically meaningful point-mutant peptides that affect binding affinity, distinguishing its ability in predicting TCR specificity of point-mutants from alternative sequence-based methods. Its application is broadly applicable to studies involving both closely related and structurally diverse TCR-peptide pairs.more » « less
-
With an increased focus on incorporating fairness in machine learning models, it becomes imperative not only to assess and mitigate bias at each stage of the machine learning pipeline but also to understand the downstream impacts of bias across stages. Here we consider a general, but realistic, scenario in which a predictive model is learned from (potentially biased) training data, and model predictions are assessed post-hoc for fairness by some auditing method. We provide a theoretical analysis of how a specific form of data bias, differential sampling bias, propagates from the data stage to the prediction stage. Unlike prior work, we evaluate the downstream impacts of data biases quantitatively rather than qualitatively and prove theoretical guarantees for detection. Under reasonable assumptions, we quantify how the amount of bias in the model predictions varies as a function of the amount of differential sampling bias in the data, and at what point this bias becomes provably detectable by the auditor. Through experiments on two criminal justice datasets– the well-known COMPAS dataset and historical data from NYPD’s stop and frisk policy– we demonstrate that the theoretical results hold in practice even when our assumptions are relaxed.more » « less
-
Link prediction has been widely applied in social network analysis. Despite its importance, link prediction algorithms can be biased by disfavoring the links between individuals in particular demographic groups. In this paper, we study one particular type of bias, namely, the bias in predicting inter-group links (i.e., links across different demographic groups). First, we formalize the definition of bias in link prediction by providing quantitative measurements of accuracy disparity, which measures the difference in prediction accuracy of inter-group and intra-group links. Second, we unveil the existence of bias in six existing state-of-the-art link prediction algorithms through extensive empirical studies over real world datasets. Third, we identify the imbalanced density across intra-group and inter-group links in training graphs as one of the underlying causes of bias in link prediction. Based on the identified cause, fourth, we design a pre-processing bias mitigation method named FairLP to modify the training graph, aiming to balance the distribution of intra-group and inter-group links while preserving the network characteristics of the graph. FairLP is model-agnostic and thus is compatible with any existing link prediction algorithm. Our experimental results on real-world social network graphs demonstrate that FairLP achieves better trade-off between fairness and prediction accuracy than the existing fairness-enhancing link prediction methods.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.3389/fimmu.2023.1288105
