A Method for the Determination of the MI Threshold in the Prediction of Specificity Determining Positions

ATENCIO, HM; ORTS, F; STOCCHI, N; BRUN M; ARJEN TEN HAVE

Buenos Aires

Conferencia; 4th International Society of Computational Biology-Latin America Conference; 2016

International Society of Computational Biology

p { margin-bottom: 0.1in; direction: ltr; color: rgb(0, 0, 10); line-height: 120%; }p.western { font-family: "Calibri",sans-serif; font-size: 11pt; }p.cjk { font-family: "Times New Roman",serif; font-size: 11pt; }p.ctl { font-family: "Times New Roman",serif; font-size: 11pt; }BackgroundUDP-glycosyltransferases(UDP-GTs) transfer a sugar moiety from a UDP activated sugar to anacceptor molecule. Related to the large acceptor variety, plants haveover 80 paralogs, UDP-GTs are a paradigm in Structure-FunctionPrediction. We identify Cluster Determining Positions and predicttheir interaction levels based on Mutual Information (MI). Then, weconsider highly interactive CDPs as Specificity Determining Positions(SDPs), biochemically defined as positions that affect proteinspecificity.MIscores and significance threshold depend on the dataset. We present amethod for the empirical determination of this threshold. PlantUDP-GTs share a conserved 44 amino acid motif denoted Plant SecondaryProduct Glycosyltransferase (PSPG) motif, involvedin UDP moiety binding. Thesepositions show high interaction levels and serve as control positions(CPs) in the determination of a functionalitythresholdfor MI scores.ResultsWedefined FunctionalEnrichment (FE)as the ratio between FunctionFraction (thenumber of network CPs divided by the total number of CPs) andBackgroundFraction (thenumber of network nodes divided by the total number of positions). Weidentified two coexisting networks that show FE of 3.61 and 3.26 atmoderate MI levels. While lowering the MI level FE remains stableuntil reaching low levels at which FE suddenly drops below 1. Variouspartial datasets were analyzed using a script in order to identifySDPs and analyze the method. ConclusionsMIscores of known important positions can be used to separatefunctionallyimportant positionsfrom functionallyunimportant positions,therewith assisting in the identification of SDPs.Supportedby INTA, CONICET and AGENCIA.