Identification and Substantiation of Specificity Determining Residue Networks using small Datasets and MI-promiscuity

ORTS, F; ATENCIO, HM; 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; line-height: 120%; }BackgroundSpecificityDetermining Positions (SDPs) are positions important in thefunctional diversification of proteins and various computationalmethods for their prediction exist. Clustering Determining Positions(CDPs), those that contribute significantly to the observedclustering, can be the result of natural selection but also ofgenetic drift. CDPs can have high Mutual Information values with manyother positions as a result of high levels of molecular interaction.A such, high cumulative MI (cMI) is indicative for co-evolution andCDPs with high cMI should be considered as SDPs. The use of MI alsoidentifies Specificity Determining Networks (SDNs), SDPs directlyconnected by means of high MI. Such apparently co-evolving SDPs arelikely related to the same functional diversification.Unfortunately,solid MI evaluation requires a large amount of sequences withrelatively high variation. We consider that a functionaldiversification that evolved within a particular clade is independentof the evolution in other clades. As such, MI evaluation ofinherently small subfamily datatsets can be inevitable. Thus,although we have shown that our method of SDP identification isfunctional (Bustamante et al., 2016), we need additional metrics inorder to substantiate predictions. Here we present the development ofSDNs together with a substantiation method for SDNs and theircomponents.ResultsThebasis for the methodology was the case study of the EukaryoticSedolisin family in which we identified three well separated clades.One clade is composed of non-fungal sedolisins, another by fungalendo-sedolisins and the third by fungal exo-sedolisins.Tree topology and the required diversification suggest independentdiversification processes occurred in the two fungal clades, which issupported by results obtained from a comparison of three MI networks(endo, exo and endo+exo). From 25 putative SDPs identified in theendo-sedolisin subfamily, eight SDPs directly connected with high MI,with each SDP connecting to in between 2 and 5 other SDPs (Figure).The relatively high connectivity among these initially putative SDPsfurther indicates these positions are likely SDPs. Moreover, theeight identified SDPs have a structural position and show adivergence of physicochemical characteristics that reasonably explainthe functional divergence. These data indicate that SDPs connectivitycan be used to substantiate SDPs and the SDNs formed. An additionalmetric can be found in the amount and quality of the connection ofthe network-SDPs with non-SDPs in the same network: Some non-SDPs areconnected to several SDPs from a network, generating an additionalsubstantiation source. On the other hand, since SDPs as well asnon-SDPs have different levels of interaction promiscuity. Hence, acorrection by means of normalization must be included.Thenext step in the methodology development is to elaborate an objectivemetric based on the mentioned connectivity characteristics. Severalmeasures were defined and tested using the sedolisin case anddatasets sufficiently large for MI evaluation without connectivitysubstantiation.