Effect of gap penalty on Local Alignment Score:Score: 161 at (seq1)[2..36] : (seq2)[53..90] 2 ASTV----TSCLEPTEVFMDLWPEDHSNWQELSPLEPSD || | | ||||||||||||||||||||||||||| 53 ASSVSVGATEA-EPTEVFMDLWPEDHSNWQELSPLEPSD Score:Score: 156 at (seq1)[10..36] : (seq2)[64..90] 10 EPTEVFMDLWPEDHSNWQELSPLEPSD ||||||||||||||||||||||||||| 64 EPTEVFMDLWPEDHSNWQELSPLEPSD Blosum 62 Gap: -15 Ex: -3 Gap: -5 Ex: - 1

Finding the most “profitable” route from a complete map MATWLI.. G1 M S11 A S22s32S42 W s33S43 T S44S54 V A Bioinformatics (Mount) p Needleman –Wunsch Algorithm Smith-Waterman Algorithm

Smith-Waterman vs. Heuristic approach MATWLI G1.. M S11.. A s12S22s32S42.. W s33S34.. T V A Smith-Waterman Algorithm Finding the best alignment based on complete calculation of route map BLAST, FASTA Try to find the best alignment based on sounding assumptions. Rule of thumb: A high similarity core Often without gap

BLAST – Basic Local Alignment Search Tool It is based on local alignment, -- highest score is the only priority in terms of finding alignment match. -- determined by scoring matrix, gap penalty It is optimized for searching large data set instead of finding the best alignment for two sequences

BLAST – Basic Local Alignment Search Tool 1.A high similarity core (2- 4aa) 2.Often without gap Query: M A T W L I. Word : M A T A T W T W L W L I 1.For each word, find matches with Score > T. 2.Extend the match as long as profitable. - High Scoring segment Pair (best local alignment) 3.Find the P and E value for HSP(s) with Score > cut off*. * Cut off value can be automatically calculated based on E

BLAST – Basic Local Alignment Search Tool The P and E value for HSP(s) : based on the total score (S) of the identified “best” local alignment. P (S) : the probability that two random sequence, one the length of the query and the other the entire length of the database, could achieve the score S. E (S) : The expectation of observing a score >= S in the target database. For a given database, there is a one to one correspondence between S and E(s) -- choosing E determines cut off score

BLAST – Basic Local Alignment Search Tool BLASTN BLASTP TBLASTN compares a protein query sequence against a nucleotide sequence database dynamically translated in all reading frames. BLASTX compares a nucleotide query sequence translated in all reading frames against a protein sequence database TBLASTX compares the six-frame translations of a nucleotide query sequence against the six-frame translations of a nucleotide sequence database. Please note that tblastx program cannot be used with the nr database on the BLAST Web page.

BLAST – Advanced options : all adjustable in stand alone BLAST -F Filter query sequence [String] default = T -M Matrix [String] default = BLOSUM62 -G Cost to open gap [Integer] default = 5 for nucleotides 11 proteins -E Cost to extend gap [Integer] default = 2 nucleotides 1 proteins -q Penalty for nucleotide mismatch [Integer] default = -3 -r reward for nucleotide match [Integer] default = 1 -e expect value [Real] default = 10 -W wordsize [Integer] default = 11 nucleotides 3 proteins -T Produce HTML output [T/F] default = F …..

Smith-Waterman vs. Heuristic approach Smith-Waterman Algorithm Finding the best alignment based on complete route map BLAST, FASTA Try to find the best alignment based on experience/knowledge Search result Difference ? - For real good matches, almost no difference - For marginal similarity and exceptional cases, the difference may matter.

Overview of homology search strategy 1.) Where should I search? NCBI Has pretty much every thing that has been available for some time Genome projects Has the updated information (DNA sequence as well as analysis result)

Overview of homology search strategy 2.) Which sequence should I use as the query? Protein cDNA Genomic Test Practice: identify potential orthologs using either cDNA or protein sequence.

Overview of homology search strategy 2.) Which sequence should I use as the query? cDNA (BlastN) Protein (TblastN)

Overview of homology search strategy 2.) Which sequence should I use as the query? Protein v.s cDNA Searching at the protein level is much more sensitive query: S A L query: TCT GCA TTG target: S A L target: AGC GCT CTA Protein: ~ 5% Nucleotide: ~ 25% Base level identity Protein: 100% Nucleotide: 33%

Overview of homology search strategy 2.) Which sequence should I use as the query? If you want to identify similar feature at the DNA level. Be Cautious with genomic sequence initiated search Low complexity region repeats

Overview of homology search strategy 3.) Which data set should I search? Protein sequence (known and predicted) blastP, Smith_Waterman Genomic sequence TblastN EST TblastN Predicted genes TblastN

Overview of homology search strategy 4.) How to optimize the search ? Scoring matrices Gap penalty Expectation / cut off Example

Overview of homology search strategy 5.) How do I judge the significance of the match ? P-value, E -value Alignment Structural / Function information

Overview of homology search strategy 6.) How do I retrieve related information about the hit(s) ? NCBI is relatively easy The scope of information collection can be enlarged by searching (linking) multiple databases (links). exampleexample genome projects often have their own interface and logistics (ie. Ensemble, wormbase, MGI, etc. )Ensemble

Overview of homology search strategy 7.) How to align (compare) my query and the hits ? Global alignment Local alignment ClustalW/ClustalX

Situations where generic scoring matrix is not suitable Short exact match Specific patterns

After Blast – which one is “real”?

Position –specific information about conserved domains is IGNORED in single sequence –initiated search BID_MOUSE SESQEEIIHN IARHLAQIGDEM DHNIQPTLVR BAD_MOUSE APPNLWAAQR YGRELRRMSDEF EGSFKGLPRP BAK_MOUSE PLEPNSILGQ VGRQLALIGDDI NRRYDTEFQN BAXB_HUMAN PVPQDASTKK LSECLKRIGDEL DSNMELQRMI BimS EPEDLRPEIR IAQELRRIGDEF NETYTRRVFA HRK_HUMAN LGLRSSAAQL TAARLKALGDEL HQRTMWRRRA Egl-1 DSEISSIGYE IGSKLAAMCDDF DAQMMSYSAH BID_MOUSE SESQEEIIHN IARHLAQIGDEM DHNIQPTLVR sequence X SESSSELLHN SAGHAAQLFDSM RLDIGSTAHR sequence Y PGLKSSAANI LSQQLKGIGDDL HQRMMSYSAH Why a BLAST match is refused by the family ?

Specific patterns 1.DNA pattern – Transcription factor binding site. 2.Short protein pattern – enzyme recognition sites. 3.Protein motif/signature.

Binary patterns for protein and DNA Caspase recognition site: [EDQN] X [^RKH] D [ASP] Examples: Observe: Search for potential caspase recognition sites with BaGua

Searching for binary (string) patterns Seq: A G G G C T C A T G A C A G R C W G A C A G T G R C W G A C A G T G R C W G A C A G T Positive match

Does binary pattern conveys all the information ? Weighted matrix / profile HMM model For searching protein domains

What determine a protein family? Structural similarity Functional conservation

Practice: motif analysis of protein sequence using ScanProsite and Pfam 1.Open two taps for Pfam, input one of the Blast hits and one candidate TNF to each data window. 2.Compare the results

Scan protein for identified motifs A service provided by major motif databases such as Prosite,, Pfam, Block, etc. Protein family signature motif often indicates structural and function property. High frequency motifs may only have suggestive value.

What is the possible function of my protein? Which family my protein belongs to? -- Profile databases Pfam ( ) Prosite ( ) IntePro ( ) Prints ( TS.html ) TS.html

Protein motif /domain Structural unit Functional unit Signature of protein family How are they defined?

How do we represent the position specific preference ? BID_MOUSE I A R H L A Q I G D E M BAD_MOUSE Y G R E L R R M S D E F BAK_MOUSE V G R Q L A L I G D D I BAXB_HUMAN L S E C L K R I G D E L BimS I A Q E L R R I G D E F HRK_HUMAN T A A R L K A L G D E L Egl-1 I G S K L A A M C D D F Binary pattern: L [GSC] [HEQCRK] X [^ILMFV] Basic concept of motif identification 2.

How do we represent the position specific preference ? BID_MOUSE I A R H L A Q I G D E M BAD_MOUSE Y G R E L R R M S D E F BAK_MOUSE V G R Q L A L I G D D I BAXB_HUMAN L S E C L K R I G D E L BimS I A Q E L R R I G D E F HRK_HUMAN T A A R L K A L G D E L Egl-1 I G S K L A A M C D D F Statistical representation G: 5 -> 71% S: 1 -> 14 % C: 1 -> 14 % Basic concept of motif identification 2.

Scoring sequence based on Model Seq: A S L D E L G D E A C D ….... position_1 = s(A/1) + s(S/2) + s(L/3) + s(D / 4) An example of position specific matrixexample

Representation of positional information in specific motif M-C-N-S-S-C-[MV]-G-G-M-N-R-R. Binary patterns: Positional matrix: