Ridge Regression Estimated Linear Probability Model Predictions of O-glycosylation in Proteins with Structural and Sequence Data

TL;DR

This study uses ridge regression to predict O-GlcNAc glycosylation in human proteins, demonstrating the importance of structural data alongside sequence information for accurate predictions.

Contribution

It introduces a linear probability model incorporating structural and sequence data to predict O-glycosylation, highlighting structural features' significance.

Findings

Structural attributes significantly predict glycosylation.

Sequence data alone yields lower prediction accuracy.

Structural data improves model performance (KS=99%).

Abstract

The likelihood of O-GlcNAc glycosylation in human proteins is predicted using the ridge regression estimated linear probability model (LPM). To achieve this, sequences from three similar post-translational modifications (PTMs) of proteins occurring at, or very near, the S or T site are analyzed: N-glycosylation, O-mucin type (O-GalNAc) glycosylation, and phosphorylation. Results found include: 1) The consensus composite sequon for O-glycosylation does NOT have W on either side of the glycosylation site. 2) The same holds for the consensus sequon for phosphorylation. 3) For LPM estimation, N-glycosylated sequences are found to be good approximations to non-O-glycosylatable sequences. 4) The selective positioning of an amino acid along the sequence, differentiates the PTMs of proteins. 5) Some N-glycosylated sequences are also phosphorylated at the S or T site. 6) ASA values for N-glycosylated sequences are stochastically larger than those for O-GlcNAc glycosylated sequences. 7) Structural attributes (beta turn II, II', helix, beta bridges, beta hairpin, and the phi angle) are significant LPM predictors of O-GlcNAc glycosylation. The LPM with sequence and structural data as explanatory variables yields a Kolmogorov-Smirnov (KS) statistic value of 99%. 8) With only sequence data, the KS statistic erodes to 80%, underscoring the germaneness of structural information, which is sparse on O-glycosylated sequences. With 50% as the cutoff probability for predicting O-GlcNAc glycosylation, this LPM mispredicts 21% of out-of-sample O-GlcNAc glycosylated sequences as not being glycosylated. The 95% confidence interval around this mispredictions rate is 16% to 26%