AndMotioninBiology

5.1.Ex:Coiledcoils

NewCoils is an algorithm for predicting coiled coils motifs that works by multiplying theprobabilities of frequencies of each amino acid residue in each coiled coil position.Animprovement upon this algorithm is PairCoil which multiplies pairwise probabilities ofspatially neighboring positions. If an amino acid is at position n, then its two neighborsare at positions n+1 and n+3 (Figure 6). The algorithm uses a sliding window of 28amino acids, determined heuristically.PairCoil works better than NewCoils because itconsiders the interactions between spatially adjacent amino acids.In fact PairCoil hasperfectscoreseparationbetweencoiledcoilalphahelicesandnon-coiledcoilalphahelices.
Figure6:Topviewofcoiledcoilmotif

Figure7:Betahelixmotif

2. 5.2.Ex:Betahelices

Thebetahelixisaprokaryoticmotifmadeupofthreebetasheets(Figure7).Currently,there is only one algorithm for predicting Beta helices, called BetaWrap. Predicting Betahelicesiscomputationallydifficultbecausethereareveryfewsolvedstructures,eachofwhichisverydifferentfromoneanother.Thealgorithmhasthreestages:


RungssubproblemGiventhelocationofaT2turnofarung,findlocationofT2turnofnextrung.


MultiplerungsFindmultipleinitialB2-T2-B3rungsandfind”optimalwrap”using dynamicprogramming.


CompletingtheparseFindB1strandsbylocallyoptimizingtheirlocation.

6. 6.SecondaryStructurePrediction

The goal of secondary structure prediction is to classify the positions of a given aminoacid sequence into alpha helices, beta strands, and loops. In general this problem isharder than motif identification.The best method for solving this problem is Neuralnetworks.One such algorithm for classifying secondary structures is PSIPRED. Thebasicstepsinthisalgorithmare:

1. 
   Givenasequencex,generateitsprofileusingPSI-BLAST(seebelow).
   
2. 
   Passtheprofiletoapre-trainedneuralnetwork.
   
3. 
   Outputclassification:alphahelix/betastrand/loop.PSIPREDhasa76%classificationaccuracy.
   

1.