Bioinformatics amp; Algorithmics.www.stats.ox.ac.ukheinlectures..ppt

Turning cabbage into a turnip Accelerations of pairwise algorithm Exact acceleration (Ukkonen,Myers). Assume all events cost 1. If de(s1,s2) 2e+|l1-l2|, then d(s1,s2)= de(s1,s2) Heuristic acceleration: Smaller band larger acceleration, but no guarantee of optimum. e { Alignment of many sequences. s1=ATCG, s2=ATGCC, ......., sn=ACGCG Alignment: AT-CG s1 s3 s4 ATGCC \ ! / ..... ---------- ..... / \ ACGCG s2 s5 Configurations in an alignment column: 2n-1 Recursion: Di=min{Di-? + d(i,?)} ? [{0,1}n\{0}n] Initial condition: D0,0,..0 = 0. Computation time: ln*(2n-1)*n Memory requirement: ln (l:sequence length, n:number of sequences) Longer Indels TCATGGTACCGTTAGCGT GCA-----------GCAT gk :cost of indel of length k. Initial condition: D0,0=0 Di,j = min { Di-1,j-1 + d(s1[i],s2[j]), Di,j-1 + g1,Di,j-2 + g2,, Di-1,j + g1,Di-2,j + g2,, } Cubic running time. Quadratic memory. (i,j) (i-1,j) (i-2,j) (i,j-1) (i,j-2) Evolutionary Consistency Condition: gi + gj gi+j If gk = a + b*k, then quadratic running time. Gotoh (1982) Di,j is split into 3 types: 1. D0i,j as Di,j, except s1[i] must mactch s2[j]. 2. D1i,j as Di,j, except s1[i] is matched with -. 3. D2i,j as Di,j, except s2[i] is matched with -. n n n n n - - n + + + n - n n n - + + - n n n - n + + 0: 1: 2: Then:D0i,j = min(D0i-1,j-1, D1i-1,j-1, D2i-1,j-1) + d(s1[i],s2[j]) D1i,j = min(D1i,j-1 + b, D0i,j-1 + a + b) D2i,j = min(D2i-1,j + b, D0i-1,j + a + b) Distance-Similarity. (Smith-Waterman-Fitch,1982) Di,j=min{Di-1,j-1 + d(s1[i],s2[j]), Di,j-1 +g, Di-1,j +g} Si,j=max{Di-1,j-1 + s(s1[i],s2[j]), Si,j-1 -w, Si-1,j-w} Distance: Transitions:2 Transversions 5 Indels:10 M largest distance between two nucleotides (5). Similarity s(n1,n2) M - d(n1,n2) wk