1 / 45

Sequence Similarity & Alignment with BLAST

Sequence Similarity & Alignment with BLAST. June 12, 2014. Outline. How sequences are aligned How alignments are scored The different BLAST algorithms Using NCBI BLAST programs. Sequence alignment. Determine if & how two sequences are related Sequence assembly Sequence annotation

rufus
Télécharger la présentation

Sequence Similarity & Alignment with BLAST

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Sequence Similarity & Alignment with BLAST June 12, 2014

  2. Outline • How sequences are aligned • How alignments are scored • The different BLAST algorithms • Using NCBI BLAST programs

  3. Sequence alignment • Determine if & how two sequences are related • Sequence assembly • Sequence annotation • Identify shared protein domains or motifs • Analysis of genomes • Phylogeny and evolution

  4. Definitions • Homologous – share a common ancestor • Cannot be measured • Measure similarity; infer homology • Orthologs: separated by speciation • Paralogs: separated by duplication

  5. Defining orthologs • Alternative splicing is estimated to occur in ~90% of human genes • Contributes an order of magnitude to the transcriptional complexity • Caspase 9 gene • Longest transcript is pro-apoptotic • Shorter transcript lacks 4 exons and a functional protease domain and is anti-apoptotic • Mouse – human transcript comparison • ~25% of the human transcripts (13% of Refseq genes) have no splicing ortholog in the mouse genome

  6. Pairwise alignments • How are two sequences related to each other? • Are there gaps in one versus the other? • What is the percent similarity? • How do I determine the significance?

  7. Pairwise alignment First string = a b c d e Second string = a c d d e f Two alignments:a b c d - e – a – c d d e f a b c – d e – a – c d d e f How do we define the criteria so that an algorithm will choose the best alignment?

  8. Defining the algorithm • Are sequences DNA or protein • What is the expected level of similarity • Scoring method that reflects degree of similarity • Allow for gaps (insertions and deletions) • Statistical measure of the probability that alignment occurred by chance

  9. % identity and similarity • % identity: percentage of aligned residues that are identical • % similarity: percentage of aligned residues that have similar chemical/physical properties • Amino acid alignments only

  10. Scoring schemes • Method of scoring matches, mismatches & gaps that is biologically relevant • Nucleotide alignments: • Identity only, with positive score for matches & negative score for mismatches • Score transitions (AnG, TnC) & transversions (purinenpyrimidine) differently • Transitions more common and more likely to be silent

  11. Amino acid substitution matrices • Method to score matches and mismatches • Based on observed frequencies of amino acid distributions and substitutions • Must model conservative nature of substitutions • Implicitly represent evolutionary patterns • Scores are based in Information Theory

  12. Scoring amino acid substitutions • Amino acids are NOT distributed evenly • Amino acids share similarity based on chemical and physical properties • Not all substitutions are equally likely due to physical/chemical constraints • i.e. L -> I is much more conservative than L -> Y vs

  13. Entropy H = information, associated with some probability p, is the base 2 logarithm of the inverse of p. Values converted to base 2 logarithms are given the unit bits. Information is described as a message of symbols. If there are n symbols and all n have an equal probability then the probability of any symbol appearing is 1/n

  14. Information Theory If all symbols are NOT equally probable, then the entropy (H) is the negative sum over all symbols (n) of the probability of a symbol (pi) multiplied by the log base 2 of the symbol (log pi) The entropy of a normal coin is therefore: -( (0.5)(-1) + (0.5)(-1) ) = 1 bit The entropy of a trick coin where heads comes up ¾ of the time is: -( (0.75)(-.415) + (0.25)(-2) ) = 0.81 bit The entropy of random DNA is: -( (0.25)(-2) + (0.25)(-2) + (0.25)(-2) + (0.25)(-2) ) = 2 bits

  15. Commonly observed substitutions: S > 0 Rarely observed substitutions: S< 0 Observed and random frequency same: S = 0 Scoring matrices S = score for amino acid pairing in the alignment qij is the observed pairing frequency of amino acids iand j. piand pj are the expected frequencies for amino acids iand j.

  16. BLOSUM62 Matrix • BLOcksSUbstitutionMatrix are based on protein alignments • Number indicates minimal percent identity between proteins in the alignment

  17. Amino acid chemical relationships

  18. Large positive; Rare amino acids Large negative; unlikely subs Near zero; no penalty for subs BLOSUM62 Matrix

  19. BLOSUM90 More positive; more negative than BLOSUM62 Based on blocks of aligned protein sequences that are at least 90% identical to another sequence in the block

  20. Choosing a matrix

  21. Gaps • Insertions can lead to gaps of varying lengths • Use 2 gap penalties: • higher penalty for opening a gap • lower penalty for extension of a gap

  22. BLAST Calculate statistical significance of matches • Build a list of words from query sequence (3 for proteins, 11 for DNA) • Evaluate each word for match using scoring matrix and discard all below threshold • Generally 50 matches per word • T value is threshold; determines sensitivity and speed of search Build word list from query sequence Find hits in database sequence Extend the hits to form HSPs

  23. Query sequence: PSATPVLICWAAG Word list: PSA ATP VLI CWA Threshold score (T): 11 Matches to PSA: Score: PSA15 PST9 PDA11 WSA 4

  24. BLAST • Find match for each word in database • Database is indexed so all possible words in all sequences is known • This search is very fast (500K words/sec) • Matches > threshold(T) are used as seed for alignments Calculate statistical significance of matches Build word list from query sequence Extend the hits to form HSPs Find hits in database sequence

  25. BLAST • Extend alignment from each word in both directions so long as score increases • These alignments are the high scoring pairs (HSPs) • Keep HSPs if score is above a given threshold Calculate statistical significance of matches Build word list from query sequence Find hits in database sequence Extend the hits to form HSPs

  26. Extending the hit Score of previous alignment (A) Score of new aligned pair Score of new alignment = + (1) p S A P S A 15 C C 9 P S A C P S A C 24 = + (2) Score of new aligned pair Score of previous alignment (B) Score of alignment (C) + = P S A C P S A C 24 Y W 2 P S A C Y P S A C W 26 = + (3) Repeat adding aligned pairs until score goes down or reach end of sequence.

  27. BLAST • Highest scoring HSPs extended in both directions as long as score > threshold • Do NOT usually get an alignment over the ENTIRE length of the sequence Combine HSPs into a gapped alignment Build word list from query sequence Find hits in database sequence Extend the hits to form HSPs

  28. Positives = 200/310 (64%) Identities = 135/310 (43%) Score = 272 bits Expect = 2e-73

  29. Significance of alignment probability that the observed match could have happened by chance P = number of matches as good as the observed one that would be expected to appear by chance in a database of the sizeprobed E = Expect value

  30. Significance of alignments • P: values between 0 and 1 • E = P x size of the database • E values range from 0 to the size of the database

  31. What is a significant alignment? • Identify a true ortholog between species • In a protein-protein alignment, E-values <10-25 • Are all the domains present in both? • Does the number of exons match? • Are the splice boundaries the same? • Similar function • Used 10-6 between C. neoformans and S. cerevisiae • Annotation (transfer annotation between species) • E-values < 10-25 fairly standard

  32. Caveats • Repetitive sequence • Regions of low complexity • Repeated motifs • Unusually high number of low abundant amino acids (i.e. cysteines)

  33. NCBI Blast homepage

  34. Nucleotide BLAST • Megablast • long alignments between very similar sequences • FASTEST • Can set percent identity for cut-off of the alignment • discontinuous megablast • find sequences similar, but not identical to the query, • more sensitive than megablast • blastn • most sensitive; shorter word query • SLOWEST • E value cutoff automatically adjusted for short query sequences

  35. Protein BLAST • BLASTP • protein query; protein DB • Finds local regions of similarity • Can define the scoring matrix • Automatic adjusts parameters for peptide searches • PSI-BLAST • very sensitive protein-protein searches • Uses PSSM and can find distant homologs • PHI-BLAST • restricted protein pattern search • Search with a query + pattern • Returns a match IF the pattern is matched

  36. Types of questions • What is the taxonomic distribution of my protein? • BLASTP against NR restricted to different taxonomic groups • Looking for it in a specific group? Restrict DB to group • What is the potential function of my protein? • BLASTP against Uniprot (best annotation) • Does the match contain a specific motif? • PHI-BLAST (BLASTP with a search for a motif pattern)

  37. Other BLAST algorithms • BLASTX • Query: Nucleotide, translated in 3 frames • DB: protein • You have an EST and want to find similarity to protein • TBLASTN • Query: protein • DB: Nucleotide, translated in 3 frames • Looking for protein homologs in an unannotated EST database • TBLASTX • Nucleotide query, nucleotide DB; both translated in 3 frames • Looking for novel sequences in error prone nucleotide query sequences • Very computer intensive

  38. Choosing the database nr/nt: Genbank+ Refseq Nucleotides + EMBL + DDBJ + PDB Excludes: HTGS, EST, GSS, STS, PAT and WGS • locate source of a sequence • find taxonomic distribution of a sequence

  39. Choosing the database • locate source of a sequence • find taxonomic distribution of a sequence

  40. Choosing the database

  41. Report options • Default is human readable with links to NCBI records • Can download a hit table in CSV format and import into Excel • Taxonomy report • Shows distribution of all hits by lineage and taxonomic group • How many results will it return? • Default is 100. May need to increase that or restrict the database to confirm a negative result

  42. Specialized BLAST • Blast2seq • Click on the “Align 2 or more sequences” and the interface changes to allow you to put in a second sequence • Can align with all the major blast programs • Get both alignment as well as a dotplot of the alignment • Useful for quick comparison of a few sequences

  43. Primer BLAST • Find primers specific to your PCR template • Can check for specificity with BLAST

  44. This weeks exercise • Using BLAST to identify the source of unknown DNA sequences • Using BLAST to identify taxonomic distribution of an unknown sequence • Using BLAST to identify homologs of specific proteins in other species • Using Primer-BLAST to find locations of primers

More Related