Efficient discovery of immune response targets by cyclical refinement of QSAR models of peptide binding

Overview of Quantitative Structure–Activity Relationships (QSAR)

“…Examples of recent work in this important area includes toxicity prediction (Cronin 2000, Freidig andHermens 2001) physicochemical properties prediction (e.g. water solubility, lipophilicity) (Beck et al 2000;Xing and Glen 2002), gastrointestinal absorption (Zhao et al 2001;AgatonovicKustrin et al 2001;Benigni et al 2000), activity of peptides (Brusic et al 2001), data mining, drug metabolism (Lewis 2000), and prediction of other pharmacokinetic and ADME properties. Recent reviews (e.g.…”

QSAR method for prediction of protein-peptide binding affinity: Application to MHC class I molecule HLA-A*0201

“…More recently, based on sequence information, a number of methods and algorithms have also been introduced to identify and characterize the T-cell epitopes, including binding motif scheme to matrix scoring schemes [22][23][24], decision trees [25], evolutionary algorithms [26], hidden Markov [27], CoMFA (comparative molecular field analysis)/CoMSIA (comparative molecular similarity indices analysis) [28], multiple regression [29] and neutral networks [30]. As far as we know, there are few works detailing the problem of epitope prediction in a quantitative way [31][32][33]. Doytchinova and Flower [28] applied CoMFA and CoMSIA analysis to perform 3D QSAR (quantitative structure-activity relationships) study on MHC/epitope binding affinity and Lin et al [29] built a linear function for predicting binding affinity of nonapeptides.…”

Predictive Bayesian neural network models of MHC class II peptide binding

“…Their Fresno method was able to rank the affinities of the peptides, and predict numerical values for their binding energies within 3-4 kJ/mol. Brusic et al used backpropagation neural networks to derive a QSAR model and identify potent HLA-A11 binders from a training set of nonamer (nine amino acid) peptides with known binding affinities [5]. Their cyclically refined models were able to identify peptides that bound but did not conform to a putative binding motif.…”