Protein Structure and Function. Application of Bioinformatics Methods - John Rigden 2014

Bioinformatics methods for studying the structure and functions of disordered proteins
Prediction of IDP functions
Correlation of disorder model and function

As follows from the considerations outlined above, comprehensive and reliable Structure/88.html">Prediction of IDP Functions is still a long way off, and only the first steps have been taken toward this goal. The next section will discuss the few Methods that may shed light on IDP functions not yet described experimentally. The functional correlation of the global disorder model (Lobley et al. 2007), sequence-based Prediction of short linear motifs using various algorithms (Davey et al. 2006; Neduva and Russell 2006), prediction of MoRFs in IDPs/IDRs (Mohan et al. 2006; Vacic et al. 2007), and the combination of sequence information with disorder (Iakoucheva et al. 2004; Radivojac et al. 2006) represent Examples of a sound approach for evaluating the functions of an unknown intrinsically disordered protein region.

To search for a relationship between the global disorder model and protein function (Lobley et al. 2007) described in standard Gene Ontology (GO) categories, Jones et al. chose a straightforward approach. It was discovered for the first time that disorder descriptors related to both Location and length correlate with functional categories associated with signal Transduction and Transcriptional Regulation. Annotations of molecular function and biological process were used. For location descriptors, A number of trends associated with GO categories were noted, such as enrichment in the middle of the protein in transcriptional regulators, DNA binding, and RNA pol II METABOLISM/31.html">Transcription factors, in the C-terminus of transcription factor activators, transcription factor repressors, and transcription factors, as well as in the N-terminus of Proteins annotated with potassium channels. For length descriptors, even more pronounced associations with function were shown than for location descriptors. Thus, disordered regions exceeding 500 residues are overrepresented in categories related to transcription, whereas shorter regions of about 50 residues or less are prevalent in metal ion-binding proteins, as well as in proteins performing ion channel and GTPase regulatory functions. The discovered associations can be used to improve the quality of protein function predictions: when applying a support vector machine-based prediction method to 26 GO categories, predictions for 11 biological process categories and 12 molecular function categories were of higher quality As a result of incorporating disorder properties. Overall, accounting for disorder markedly improves the quality of protein function predictions, with more significant improvements observed for biological process categories compared to molecular functions.