From Protein Structure to Function with Bioinformatics.pdf

310 I.A. Cymerman et al.An early example of work in this area was modelling of brain lipid binding protein(BLBP), based on the related fatty-acid binding protein structures (Xu et al.1996). Interactions of known fatty acid ligands of BLBP were modelled in an effortto discover the molecular basis of the 20-fold tighter binding of docosahexaenoicacid relative to the shorter oleic and arachidonic acids. The model revealed that thetwo extra carbon atoms of the former fatty acid could be accommodated in thepocket of BLBP, making additional favourable hydrophobic interactions. The calculatedadditional binding free energy, based on the size of the additional hydrophobiccontact area, of around 2 kcal/mol correlated nicely with the difference inaffinity. With the model validated in this way, the authors were able to predict thatstill larger fatty acids would not be able to make additional contacts and wouldtherefore not bind any more tightly.The molecular bases of different specificities may sometimes be surprisinglysimple. Such is the case with the phospho donors of some 6-phosphofructokinases(PFKs). PFK is a glycolytic enzyme catalysing the transfer of a phospho groupfrom a donor, which may be ATP, ADP or inorganic pyrophosphate (PPi). The ATPandPPi-dependent enzymes share an evolutionary relationship, while ADPdependentPFKs belong to a different structural class. It was noticed early on thatthe ATP-dependent enzymes from trypanosomatids bore a closer relationship tocertain PPi-dependent enzymes that they did to the better-characterised ATPdependentenzymes from bacteria and mammals (Michels et al. 1997). Modellinglater revealed that the basis for ATP or PPi specificity could be pinned down to asingle amino-acid which was Gly in the ATP enzymes but Asp in the PPi enzymes(Lopez et al. 2002). As shown in Fig. 12.5, an Asp at this position clashes stericallyand electrostatically with the α-phosphate of bound ADP or ATP, reducing thebinding site to a size that can only accommodate PPi as phospho donor. The conversionof a PPi-dependent enzyme to an ATP-dependent one by the replacement ofthe Asp at this position with a Gly confirms the dramatically simple origin of specificityin this case (Chi and Kemp 2000).12.4.6 Structure Modelling of Alternatively Spliced IsoformsMany, if not most, eukaryotic genes are alternatively spliced, dramatically increasingthe diversity of transcripts. It is often difficult to predict from the sequences ofalternatively spliced transcripts whether function is retained or modified. Structuremodelling, where possible, can shed light on the structure-function relationshipamong alternatively spliced transcripts from a single gene.Early work by Furnham et al. (2004), involving 40 splice variant models of 14 proteins,showed that exon loss often involved loss of complete structural units rather thansmall regions. The authors showed that deletions were more reliably modelled, accordingto structure validation software, than insertions. For four proteins with biomedicalimplications the authors could correlate known function properties of splice variantswith their modelled structures. Later Wang et al. (2005) showed that boundaries of