Factor analysis scales of generalized amino acid information as applied in predicting interactions between the human amphiphysin-1 SH3 domains and their peptide ligands.
ABSTRACT Factor analysis scales of generalized amino acid information (FASGAI) involving hydrophobicity, alpha-helix and beta-turn propensities, bulky properties, compositional characteristics, local flexibility, and electronic properties, was proposed to represent the structures of the decapeptides binding the human amphiphysin-1 SH3 domains. Parameters being responsible for the binding affinities were selected by genetic algorithm, and a quantitative structure-affinity relationship (QSAR) model by partial least square was established to predict the peptide-SH3 domain interactions. Diversified properties of the residues between P(2) and P(-3) (including P(2) and P(-3)) of the decapeptide (P(4)P(3)P(2)P(1)P(0)P(-1)P(-2)P(-3)P(-4)P(-5)) may contribute remarkable effect to the interactions between the SH3 domain and the decapeptide. Particularly, electronic properties of P(2) may provide relatively large positive contributions to the interactions, and reversely, hydrophobicity of P(2) may be largely negative to the interactions. These results showed that FASGAI vectors can well represent the structural characteristics of the decapeptides. Furthermore, the model obtained, which showed low computational complexity, correlated FASGAI descriptors with the binding affinities as well as that FASGAI vectors may also be applied in QSAR studies of peptides.
- SourceAvailable from: plosone.org[show abstract] [hide abstract]
ABSTRACT: Bioactive peptides and peptidomimetics play a pivotal role in the regulation of many biological processes such as cellular apoptosis, host defense, and biomineralization. In this work, we develop a novel structural matrix, Index of Natural and Non-natural Amino Acids (NNAAIndex), to systematically characterize a total of 155 physiochemical properties of 22 natural and 593 non-natural amino acids, followed by clustering the structural matrix into 6 representative property patterns including geometric characteristics, H-bond, connectivity, accessible surface area, integy moments index, and volume and shape. As a proof-of-principle, the NNAAIndex, combined with partial least squares regression or linear discriminant analysis, is used to develop different QSAR models for the design of new peptidomimetics using three different peptide datasets, i.e., 48 bitter-tasting dipeptides, 58 angiotensin-converting enzyme inhibitors, and 20 inorganic-binding peptides. A comparative analysis with other QSAR techniques demonstrates that the NNAAIndex method offers a stable and predictive modeling technique for in silico large-scale design of natural and non-natural peptides with desirable bioactivities for a wide range of applications.PLoS ONE 01/2013; 8(7):e67844. · 3.73 Impact Factor
- [show abstract] [hide abstract]
ABSTRACT: Many protein-protein interactions are mediated by a peptide-recognizing domain, such as WW, PDZ, or SH3. In the present study, we describe a new method called position-dependent noncovalent potential analysis (PDNPA), which can accurately characterize the nonbonding profile between the human endophilin-1 Src homology 3 (hEndo1 SH3) domain and its peptide ligands and quantitatively predict the binding affinity of peptide to hEndo1 SH3. In this procedure, structure models of diverse peptides in complex with the hEndo1 SH3 domain are constructed by molecular dynamics simulation and a virtual mutagenesis protocol. Subsequently, three noncovalent interactions associated with each position of the peptide ligand in the complexed state are analyzed using empirical potential functions, and the resulting potential descriptors are then correlated with the experimentally measured affinity on the basis of 1997 hEndo1 SH3-binding peptides with known activities, using linear partial least squares regression (PLS) and the nonlinear support vector machine (SVM). The results suggest that: (i) the electrostatics appears to be more important than steric properties and hydrophobicity in the formation of the hEndo1 SH3-peptide complex; (ii) P(-4) of the core decapeptide ligand with the sequence pattern P(-6)P(-5)P(-4)P(-3)P(-2)P(-1)P(0)P(1)P(2)P(3) is the most important position in terms of determining both the stability and specificity of the architecture of the complex, and; (iii) nonlinear SVM appears to be more effective than linear PLS for accurately predicting the binding affinity of a peptide ligand to hEndo1 SH3, whereas PLS models are straightforward and easy to interpret as compared to those built by SVM.Journal of Molecular Modeling 09/2011; 18(5):2153-61. · 1.98 Impact Factor
- [show abstract] [hide abstract]
ABSTRACT: Peptides play significant roles in the biological world. To optimize activity for a specific therapeutic target, peptide library synthesis is inevitable; which is a time consuming and expensive. Computational approaches provide a promising way to simply elucidate the structural basis in the design of new peptides. Earlier, we proposed a novel methodology termed HomoSAR to gain insight into the structure activity relationships underlying peptides. Based on an integrated approach, HomoSAR uses the principles of homology modeling in conjunction with the quantitative structural activity relationship formalism to predict and design new peptide sequences with the optimum activity. In the present study, we establish that the HomoSAR methodology can be universally applied to all classes of peptides irrespective of sequence length by studying HomoSAR on three peptide datasets viz., angiotensin-converting enzyme inhibitory peptides, CAMEL-s antibiotic peptides, and hAmphiphysin-1 SH3 domain binding peptides, using a set of descriptors related to the hydrophobic, steric, and electronic properties of the 20 natural amino acids. Models generated for all three datasets have statistically significant correlation coefficients (r(2) ) and predictive r2 (rpred 2) and cross validated coefficient ( qLOO 2). The daintiness of this technique lies in its simplicity and ability to extract all the information contained in the peptides to elucidate the underlying structure activity relationships. The difficulties of correlating both sequence diversity and variation in length of the peptides with their biological activity can be addressed. The study has been able to identify the preferred or detrimental nature of amino acids at specific positions in the peptide sequences. © 2013 Wiley Periodicals, Inc.Journal of Computational Chemistry 09/2013; · 3.84 Impact Factor