| Humans encounter conditions of low oxygen (hypoxia) under various physiological and pathological conditions. Oxygen-dependent post-translational modifications (PTMs) can affect both the physical and chemical properties of proteins. Thus, these modifications are key elements to the cellular response to hypoxia. Asparagine or aspartate hydroxylation are examples of oxygen-dependent PTMs, and play an important role in cellular oxygen sensing. Experimental approaches to study asparagine or aspartate hydroxylation of proteins are expensive and time-consuming. Therefore, there is a need for computational prediction tools to narrow and rank potential protein targets of oxygen-dependent PTMs for experimental validation. Following a detailed review of the literature, we found 40 experimentally confirmed human protein targets with 61 positive sites for asparagine/aspartate hydroxylation. Utilizing the RVP-NET tool to predict solvent accessibility, it was found that almost all positive sites were located on the protein surface. An analysis of predicted protein secondary structure, found that all positive sites occurred in non-regular or beta-strand secondary structures. A sequence logo tool, which displays positional information of aligned amino acid sequences, found that the sequence windows surrounding each positive site showed some pattern of sequence conservation. We used a support vector machine to train a classifier on the positive sites using the input features of surface accessibility, secondary structure, and position specific scoring matrix (PSSM). This trained classifier can recall 92.73% asparagine/aspartate hydroxylation sites with 61.45% precision based on leave-one-out test method. The prediction tool - HYPT is currently available at http://bioinf.sce.carleton.ca/HYPT. |
