A Bioinformatics Approach To The Structural And Functional Analysis Of .

A bioinformatics approach tothe structural and functionalanalysis of the glycogenphosphorylase protein familyJieming Shen1,2 and Hugh B. Nicholas, Jr.31Bioengineeringand Bioinformatics Summer Institute, Department of Computational Biology,University of Pittsburgh, Pittsburgh, PA 152612Department of Chemistry and Chemical Biology, Rutgers University, Piscataway, NJ 088543Biomedical Initiative Group, Pittsburgh Supercomputing Center, Pittsburgh, PA 15213

Objectives Build a phylogenetic profile for the glycogenphosphorylase protein family from existingsequences Identify orthologues/paralogues; orthologues have samebiochemical/physiological roleProteins involved in the same pathway usually havesimilar phylogenetic profilesAids in identification of pathways and physiologicalprocesses in which uncharacterized protein appearsConduct a validation study of the ProbConsalignment algorithm Prepare multiple sequence alignments in parallel usingProbCons as well as more established methods forobtaining alignment (ClustalW)Compare .MeinelATmolgen.mpg.de 323.html

Background Information Glycogen phosphorylase plays a major role incarbohydrate metabolism by catalyzing thebreakdown of glycogen into glucose subunits Acts on linear chains of glycogenGlycogen shorted by one residueDimer of 2 identical ib/media portfolio/13.html

Methodology Overview Sequence Retrieval(iProclass)PatternIdentification andMatching (MEME)Multiple SequenceAlignment identificationRefined Multiple Sequence Alignmentand matchingMultiplePrinciple CompBootstrap Phylosequenceonent Analysisgenetic Analysisalignment(SeqSpace)(PHYLIP suite)PrinciplecomponentIdentified SubgroupsanalysisGroup EntropyPhylogeneticAnalysis (GEnt)analysisGroup entropyDistinctive Residuesanalysis3D visualizationPhylogenetic Profile

Sequence Retrieval iProclass databaseObtained 355 glycogen phosphorylasesequences of about 800 residues each Eliminated duplicates or nearduplicates—left with 282 sequences Final dataset included top mostsequenced genomes: honeybee,chicken, sea squirt, cow, dog,drosophila, Japanese puffer fish, human,mosquito, mouse, and C. elegans

Pattern Identification andMatching MEME (Multiple EM for MotifElicitation)Discovers motifs (highly conservedsequence patterns) in group of relatedprotein sequences Sorts through sequences and finds andreports motif alignments regardless oftheir placement along the protein Obtained 20 highly conserved motifs Conserved residues and motifs essentialto protein structure and function

Multiple Sequence Alignment ClustalW ProbCons Consistency-based method 10 hoursT-COFFEE, another consistency-basedmethod, was attempted but did notcomplete the MSA within a reasonableamount of time (2 weeks) Progressive method 1 hourmay have had to do with large size of datasetand long sequence lengthGeneDoc program used to highlight MEMEpatterns in the MSAs

Multiple Sequence Alignment GeneDoc: ProbCons conserved residues

Multiple Sequence Alignment GeneDoc: ClustalW conserved residues

Multiple Sequence Alignment GeneDoc: ProbCons with highlighted MEME motifs

Multiple Sequence Alignment GeneDoc: ClustalW with highlighted MEME motifs

Refined MSA Incorporated information from boththe ClustalW and ProbCons outputsRefined alignment by hand usingMEME motifs as guide

Principle Component Analysis SeqSpace One method to identify groups of subfamilies“Top-down” analysisTakes distance measures among set ofsequences and converts into self-consistent setof coordinates in some arbitrary number ofdimensions Coordinates plotted and examined for clusters ofsequencesClustered protein sequences share similar pattern ofsubstitutions—presumed to reflect some commonbiochemical/physiological property or function

SeqSpace: Dimensions 2 x 3VariousbacteriaandarchaeaMetazoa

Bootstrap PhylogeneticAnalysis with PHYLIP suite ClustalW alignment with gap columnseliminatedGblocks: eliminate poorly aligned positionsand divergent regions of sequence ofprotein alignmentSeqboot: generate multiple data sets(1000) that are resampled versions of theinput data set, randomly sample columnswith replacementProtdist: analyzes the multiple data setsand computed a distance measure forprotein sequences, using maximumlikelihood estimates based on the DayhoffPAM matrix in this case

Bootstrap PhylogeneticAnalysis with PHYLIP suiteNeighbor: constructs a tree bysuccessive clustering of lineages,setting branch lengths as the lineagesjoin Consense: computes consensus treesby majority-rule method Phylogenetic consensus tree viewedin ATV viewer and TreeView Groups of subfamilies compared tothose obtained from Seqspace forfurther refinement

TreeView Unrooted tree forMetazoa subfamilyrecalculated fromdistinct subset ofoverall treeReveals enzymeisozyme groups forbrain, muscle, andliver tissue withinthe moremedi.htmhttp://en.wikipedia.org/wiki/Biceps brachii

GEnt Group Entropy Analysis Identify distinctive features of mutuallyexclusive subsets determined byphylogenetic analysis through crossentropy analysis of the columns in MSAWithin single column, contrast: Amino acid composition within defined group ofsequencesAmino acid composition in all sequencesoutside of groupAlignment positions where residuecomposition inside groups is very differentfrom that outside group are expected toindicate positions associate with distinctiveproperties of sequences of that subgroup

Group Entropy Equations pi Family Entropy Distance ( pi ) log 2 qi pi Group Entropy Distance ( pi qi ) log 2 qi pi foreground residue frequencyqi background residue frequency

GEnt Results for MetazoaSubfamily5.05.5Important to family3.53.02.52.0Family y-w360k-f0.5479f-fNot enoughinformation02468Group Entropy1012Importanttosubfamily1416

2w-w1.0322w-w1304d-a1141k-l249d-q1223l-t0.5Family Entropy4.04.55.05.5GEnt Results for FungiSubfamily0123456Group Entropy7891011

GEnt Distinctive subfamily residues highlighted

Visualization3D structures for human liver isoformand yeast obtained from Protein DataBank Residues of interest highlighted inVMD

Visualization: Human LiverIsoformGEnt: distinctive residuesUniProt:binding/association sites

Visualization: Human LiverIsoform Globally conserved MEME motifs

Visualization: YeastGEnt: distinctive residuesUniProt: binding sites

Visualization: Yeast Globally conserved MEME motifs

Conclusions GEnt In metazoans: identified residue (155) knownto be important in allosteric controlOther predicted residues have high probabilityof reflecting important adaptations of enzymeto environmentDetermined possible evolutionary path ofglycogen phosphorylaseMSA methodology ProbCons and T-COFFEE did not perform welldue to the large number of sequences and longsequence length of the datasetIn this case, the ClustalW alignment methodgave the best results

Future WorkExperimentally test GEnt-predictedresidues (site-directed mutagenesis) Additional computation work withquantum mechanics and molecularmechanics to identify roles ofconserved residues, motifs, and GEntpredicted residues

AcknowledgementsHugh B. Nicholas, Jr. Alexander J. Ropelewski Ricardo G. Mendez Pittsburgh Supercomputing Center Bioengineering & BioinformaticsSummer Institute

References PSC website ials/)Michael M. Crerar, lm.htm)Buchbinder et al. (2001) Structuralrelationships among regulated andunregulated phosphorylases. Annu. Rev.Biophys. Biomol. Struct. 30: 191-209.Nicholas Jr., H.B. Glutathione S-transferasesubfamily differences: remodeling thesubunit and domain interfaces.

A bioinformatics approach to the structural and functional analysis of the glycogen phosphorylase protein family Jieming Shen1,2 and Hugh B. Nicholas, Jr.3 1Bioengineering and Bioinformatics Summer Institute, Department of Computational Biology, University of Pittsburgh, Pittsburgh, PA 15261