1. GNANA SUNDAR RAJENDIRAN JOYESH MISHRA RISHI MISHRA FALL 2008 BIOINFORMATICS Clustering Method for Repeat Analysis in DNA sequences

2. Abstract Implement a Proposed Clustering Technique by Volfovsky et al. for Repeat Analysis Distribute Merging Repeats into Classes/Clusters based on Similarity Measures Why analyze repeats?

3. Algorithm Description Selection using RepeatMatch Steps 1. Preprocessing 2. Merging and Repeat Map Generation 3. Classification 4. BLAST searches & further merging of Clusters Results

4. RepeatMatch Uses Suffix Tree algorithm to determine all the exact repeats in a given sequence It is a part of the MUMmer package used for the rapid alignment of very large DNA and amino acid sequences.

5. Example… Forward & Reverse Complement Repeats

6. Definitions Exact Repeat: Subsequence that occurs in DNA at least twice.  Exact repeat is represented by pair of co-ordinates (A1,A2) delimiting its location in the genome sequence and by the repeat length. Maximal Repeat: Repeat that can not be extended in either direction without incurring a mismatch Initial Repeat Set: Set of repeats chosen initially, from which the repeat classes will be constructed

7. Definitions

8. Preprocessing Output from RepeatMatch is used to partition the original genome sequence Each partition point has a reference to the pair co- ordinates (A1, A2) and repeat length l For each repeat starting at co-ordinates A1 and A2, with length l, this list will include both (A1, A2, l) and (A2, A1,l)

9. Merging & Repeat Map Generation This procedure works by repeatedly working together two exact repeats that either overlap or that occur within a limited distance(a gap) of each other. Significant subsequences of the new merging repeats appear at least twice in the genome sequence.  Merging with Gap  Merging with Overlap

10. Merging with Gap Given 2 partition points: p 1 = (A 1, A 2, l A ) and p 2 = (B 1, B 2, l B ) [A 1 < B 1 ] Compute the distance between the non-overlapping repeats as d(p 1,p 2 ) = max(0, B 1 - A 1 - l A + 1) Given a maximum Gap size G > 0 Sequences corresponding to p 1 and p 2 are merged if d(p 1, p 2 ) < G

11. Merging with Overlap Merges sequences which are partially identical Overlap of 2 sequences is denoted as: o(p 1, p 2 ) = max(o, A 1 + l A – B 1 + 1) for A 1 <B 1 Given a minimum overlap proportion op, where 0 <= op <= 1, repeat points (A 1, A 2, l A ) and (B 1, B 2, l B ) are merged if at least one of the four repeats has overlap satisfying o(p 1, p 2 ) > op min(l A, l B ) op : Interpreted as a fraction of the shorter of the 2 repeats. E.g. if op = 0.75, the two overlapped sequences are merged if the length of their overlap is at least 75% of the length of the shorter sequence.

12. Output: Merging Repeat The new sequence is defined as merging repeat with starting position M = A 1 and with length l M = max(A 1 + l A, B 1 + l B ) – A 1 A data structure stores with each merging repeat its start co-ordinate, the length(n M ), and a list of references to itself and to other repeats.

13. Classification If a merging repeat has at least one reference in common with one another, then they belong to the same class If a merging repeat has references that belong to multiple distinct classes, then those classes are combined into one. If a merging repeat contains no reference to an existing class, then the merging repeat forms a new class.

14. BLAST searches and further merging To merge similar but non-exact repeats Search all merging repeats against all others  A local BLAST database is created with all repeat sequences using formatdb Classes are merged if any of their underlying sequences have a BLAST E-value less than a User- Specified Threshold when compared to any sequence in another class If a class appears in multiple similarity pairs, all these similar classes are merged with the original class

15. Results

16. Results (contd.)

22. What’s Next? The results collected help to cluster similar repeats and generate a database. This DB, in future, can be used to find Similar Repeats faster. Also, if on future classification, we find common traits among repeats, they can be analyzed only for that cluster alone.

23. Resources Softwares Used:  RepeatMatch (MUMmer 3.0)  BLAST (formatdb, blastall)  MS Excel 2007 Programmming Language Used: C++, UNIX Shell Script Source for Project & Presentation  www.cise.ufl.edu/~jmishra/bioinformatics.html www.cise.ufl.edu/~jmishra/bioinformatics.html

24. References Volfovsky N., Haas Brian J. and Salzberg Steven L. A clustering method for repeat analysis in DNA sequences, 2001 Mummer software : http://mummer.sourceforge.net/http://mummer.sourceforge.net/ http://www.ncbi.nlm.nih.gov/staff/tao/URLAPI/  blastall, formatdb www.ncbi.nlm.nih.gov/BLAST/

25. References http://www.tigr.org http://www.animalgenome.org/blast/docs/blast_faq. html http://www.animalgenome.org/blast/docs/blast_faq. html Of course, Wikipedia … Data obtained from: ftp://ftp.ncbi.nih.gov/genomes/Bacteria/

26. Thank You