Chapter 9MotivationFirst StepArtificial Neural NetworkDangerProfile network from HeiDelbergPHDThreading3D-1D Matching (Bowie et al.)3D-1DMethods using 3D interactions3D interactionsPotentials of mean force (POMF)Multiple Sequence ThreadingExampleSequence-Structure AlignmentEvaluating MethodsChapter 9Chapter 9Structure PredictionStructure PredictionMotivationMotivationGiven a protein, can you predict molecular structureWant to avoid repeated x-ray crystallography, but want accuracyYou could use nucleotide alignment, but what do you do with the gapped regions?More complex methods are only justified if they can be shown to perform better than simpler methodsSimpler methods are only justified if they can perform better than basic sequence alignmentGiven a protein, can you predict molecular structureWant to avoid repeated x-ray crystallography, but want accuracyYou could use nucleotide alignment, but what do you do with the gapped regions?More complex methods are only justified if they can be shown to perform better than simpler methodsSimpler methods are only justified if they can perform better than basic sequence alignmentFirst StepFirst StepSome structure comparison methods use secondary structures of the new sequencePredict location of secondary structure elements along the protein’s backbone and the degree of residue burialSupervised learning has been shown to perform well in this taskSome structure comparison methods use secondary structures of the new sequencePredict location of secondary structure elements along the protein’s backbone and the degree of residue burialSupervised learning has been shown to perform well in this taskArtificial Neural NetworkArtificial Neural NetworkPredictsStructure at this pointPredictsStructure at this pointDangerDangerYou may train the network on your training set, but it may not generalize to other dataPerhaps we should train several ANNs and then let them vote on the structureYou may train the network on your training set, but it may not generalize to other dataPerhaps we should train several ANNs and then let them vote on the structureProfile network from HeiDelbergProfile network from HeiDelbergfamily (alignment is used as input) instead of just the new sequenceOn the first level, a window of length 13 around the residue is used The window slides down the sequence, making a prediction for each residueThe input includes the frequency of amino acids occurring in each position in the multiple alignment (In the example, there are 5 sequences in the multiple alignment)The second level takes these predictions from neural networks that are centered on neighboring proteins The third level does a jury selectionfamily (alignment is used as input) instead of just the new sequenceOn the first level, a window of length 13 around the residue is used The window slides down the sequence, making a prediction for each residueThe input includes the frequency of amino acids occurring in each position in the multiple alignment (In the example, there are 5 sequences in the multiple alignment)The second level takes these predictions from neural networks that are centered on neighboring proteins The third level does a jury selectionPHDPHDPredicts 4Predicts 4Predicts 6Predicts 6Predicts 5Predicts 5ThreadingThreadingThreading matches structure to sequenceTrue threading considers 3D spatial interactionsThreading matches structure to sequenceTrue threading considers 3D spatial interactions3D-1D Matching (Bowie et al.)3D-1D Matching (Bowie et al.)Convert 3D structure into a stringInclude -helix, -sheet or neitherInclude buried or solvent accessible (6 levels) Total of 3X6=18 distinct statesWith Pa:j= probability of finding amino acid (a) in environment (j) and Pa=probability of finding (a) anywhereConvert 3D structure into a stringInclude -helix, -sheet or neitherInclude buried or solvent accessible (6 levels) Total of 3X6=18 distinct statesWith Pa:j= probability of finding amino acid (a) in environment (j) and Pa=probability of finding (a) anywhere€ saj= logPa: jPa ⎛ ⎝ ⎜ ⎞ ⎠ ⎟3D-1D3D-1DCalculate the information values score on a training set of multiple alignments and the score was used as a profile for each columnWhen applied to the globin family an clearly identified myoglobins from nonglobins but not from other globinsCalculate the information values score on a training set of multiple alignments and the score was used as a profile for each columnWhen applied to the globin family an clearly identified myoglobins from nonglobins but not from other globinsMethods using 3D interactionsMethods using 3D interactionsResidues that have large separation in the sequence may end up next to each other when the protein is folded.Define a measure of contact between residues (two atoms within 5Å) and count frequency of contact between all pairs in PDBUse measure in alignment to evaluate cost, or to select the best alignment Residues that have large separation in the sequence may end up next to each other when the protein is folded.Define a measure of contact between residues (two atoms within 5Å) and count frequency of contact between all pairs in PDBUse measure in alignment to evaluate cost, or to select the best alignment3D interactions3D interactionsQuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.Potentials of mean force (POMF)Potentials of mean force (POMF)Since the notion of contact is somewhat arbitrary, a more general formulation can be triedDerive an empirical function for the propensity of each of the 400 pairs of residues to be any given distance apart.Since the notion of contact is somewhat arbitrary, a more general formulation can be triedDerive an empirical function for the propensity of each of the 400 pairs of residues to be any given distance apart.Multiple Sequence ThreadingMultiple Sequence ThreadingMultiple Sequence AlignmentAlign the most similar to create a consensus sequenceAlign consensus sequences to create overall alignmentUse the same strategy with structuresAssume that conserved hydrophobic positions should pack in the coreThis appears to be work in progress (1997)Multiple Sequence AlignmentAlign the most similar to create a consensus sequenceAlign consensus sequences to create overall