Gaussian Process: A Promising Approach for the Modeling and Prediction of Peptide Binding Affinity to MHC Proteins

Abstract

On the basis of Bayesian probabilistic inference, Gaussian process (GP) is a powerful machine learning method for nonlinear classification and regression, but has only very limited applications in the new areas of computational vaccinology and immunoinformatics. In the current work, we present a paradigmatic study of using GP regression technique to quantitatively model and predict the binding affinities of over 7000 immunodominant peptide epitopes to six types of human major histocompatibility complex (MHC) proteins. In this procedure, the sequence patterns of diverse peptides are characterized quantitatively and the resulting variables are then correlated with the experimentally measured affinities between different MHC and their peptide ligands, by using a linearity- and nonlinearity-hybrid GP approach. We also make systematical comparisons between the GP and two sophisticated modeling methods as partial least square (PLS) regression and support vector machine (SVM) with respect to their fitting ability, predictive power and generalization capability. The results suggest that GP could be a new and effective tool for the modeling and prediction of MHC-peptide interactions and would be promising in the field of computer-aided vaccine design (CAVD).

Rational design and improvement of the dimerization‐disrupting peptide selectivity between ROCK‐I and ROCK‐II kinase isoforms in cerebrovascular diseases

“…The conserved residues (including D2, W4, L5, E6, C7, I8, I11, V12, and E14) can form effective nonbonded interactions with both the ROCK‐I and the ROCK‐II; most of these interactions are nonspecific hydrophobic contacts and salt bridges, thus primarily conferring stability (but not specificity) to the kinase‐peptide complexes. The non‐conserved residues (including T1, D9, V10, and Y13) play a crucial role in defining the peptide selectivity for ROCK‐I over ROCK‐II, since they can only effectively interact with ROCK‐I (but not ROCK‐II) through specific hydrogen bonds, salt bridges, and π‐π stackings, as revealed by statistical modeling of protein‐peptide recognition . In addition, the other G3 residue cannot interact with the two kinase isoforms, which does not contribute either stability or specificity to the peptide.…”

High‐throughput statistical screening of antiinfective peptides from natural antibacterial protein repertoire: Chemometric prediction, molecular modeling, and susceptibility analysis

“…The AIP sample set was randomly split into a calibration set (~2/3) and a verification set (~1/3) . Principal component analysis SVM regression model was built based on the calibration set, and 5‐fold cross‐validation was carried out to measure internal stability and reliability of the model, which was then extrapolated to the verification set to estimate its external predictive power and generalization ability …”

Chemometrics‐based identification, characterization, and modeling of the intermolecular interaction between human osteosarcoma SH3 domain and its peptide ligands

“…At the same time, the amount of data is much greater than what was previously available. Thus, it solves the problem that mixing data from different sources without further standardization, and provides tool developers with a benchmark that aids in the generation and evaluation of prediction tools . The experimental affinities of these peptides were expressed as K d values ranging between micromolar and nanomolar levels; we herein converted the values to logarithmic form p K d to carry out statistical regression modeling.…”