Logistic regression models to predict solvent accessible residues using sequence- and homology-based qualitative and quantitative descriptors applied to a domain-complete X-ray structure learning set

Abstract

A working example of relative solvent accessibility (RSA) prediction for proteins is presented. Novel logistic regression models with various qualitative descriptors that include amino acid type and quantitative descriptors that include 20-and six-term sequence entropy have been built and validated. A domain-complete learning set of over 1300 proteins is used to fit initial models with various sequence homology descriptors as well as query residue qualitative descriptors. Homology descriptors are derived from BLASTp sequence alignments, whereas the RSA values are determined directly from the crystal structure. The logistic regression models are fitted using dichotomous responses indicating buried or accessible solvent, with binary classifications obtained from the RSA values. The fitted models determine binary predictions of residue solvent accessibility with accuracies comparable to other less computationally intensive methods using the standard RSA threshold criteria 20 and 25% as solvent accessible. When an additional non-homology descriptor describing Lobanov-Galzitskaya residue disorder propensity is included, incremental improvements in accuracy are achieved with 25% threshold accuracies of 76.12 and 74.79% for the Manesh-215 and CASP(8+9) test sets, respectively. Moreover, the described software and the accompanying learning and validation sets allow students and researchers to explore the utility of RSA prediction with simple, physically intuitive models in any number of related applications.

Exploring the limitations of biophysical propensity scales coupled with machine learning for protein sequence analysis

“…For simplicity, this problem is usually structured as a binary classification task in which Bioinformatics methods aim at predicting the exposed/buried state of each residue in the protein sequence. Residues with more than 25% of relative exposure to solvent belongs to the first class and all the others to the second 4 . For more details about this task, see 4,5 .…”

“…Residues with more than 25% of relative exposure to solvent belongs to the first class and all the others to the second 4 . For more details about this task, see 4,5 .…”

“…For this reason, finding the optimal feature representation for each problem is a central issue for bioinformaticians, and improvements in both ML models and the features used have contributed to a great deal of positive results in the field, such as the prediction of protein Secondary Structure 1–3 , Relative Solvent Accessibility 4,5 , Contact Maps 6,7 , homology 8,9 , cysteine oxidation states 10,11 , early folding events 12 , the deleteriousness of variants 13–15 and many more protein-related aspects.…”

Protein Solvent-Accessibility Prediction by a Stacked Deep Bidirectional Recurrent Neural Network

“…Fan et al [ 22 ] used gradient boosted regression trees to predict rASA and achieved a state-of-the-art performance, which is 9.4% MAE and 0.73 Pearson’s correlation coefficient (PCC) on the CB502 dataset. Another benchmark dataset Manesh215 [ 25 ] is also widely used by researchers [ 10 , 13 , 14 , 22 , 24 , 26 ] to validate prediction methods. Table 1 summarizes the recent developments in predicting the values of rASA.…”

Predicting Switch-Like Behavior in Proteins Using Logistic Regression on Sequence-Based Descriptors

“…Applying the aforementioned descriptors, our binary event describes whether or not a protein chain residue in question has switch-like characteristics. Using logistic regression with proteins is not new, and was done previously by the Lustig group predicting from sequence which R groups were on the inside of a folded protein [17] as well as residues of one protein interacting with residues on another.…”