Transcript Slide 1
Capstone Presentation Motif Discovery from Large Number of Sequences: A Case Study with Disease Resistance Genes in Arabidopsis thaliana by Irfan Gunduz I.U. School of Informatics 04/25/04 INTRODUCTION Motifs • Highly conserved regions across a subset of proteins that share the same function >Seq A >Seq B >Seq C >Seq D YNEDSKH YDDDSNH YDNDSNH YENDSKH • Motifs can be used to predict A molecule’s function A Structural Feature Family membership I.U. School of Informatics INTRODUCTION Current motif finding soft-wares: • MEME • PROSITE • PRATT, etc Do they work with large number of sequences? • Pattern discovery relies on statistical or combinatorial techniques, looking for signals • Signal-to-noise ratio becomes less clear as the number of sequences increases What to do? I.U. School of Informatics Objective Develop a computational procedure to find functional motifs from large number of sequences I.U. School of Informatics COMPUTATIONAL PROCEDURE Tools BLAST (Sequence alignment tool) BAG ( Sequence Clustering package) CLUSTAL W (Multiple sequence alignment) HMMERII (HMM based software) BLOCK MAKER (Block/Motif finder) LAMA (Block comparison tools) PERL I.U. School of Informatics COMPUTATIONAL PROCEDURE 1- Collecting and Clustering Sequences Extract well-annotated sequences of interest from genome of interest All to all pair wise comparison using Blast Estimate the best bit score for clustering Cluster sequences using BAG I.U. School of Informatics COMPUTATIONAL PROCEDURE 2 - ENRICHMENT Align multiple sequences in each cluster Start HMM based programs build profile for each cluster Search genome of interest with new profile and extract more sequences if available I.U. School of Informatics COMPUTATIONAL PROCEDURE 3 – REFINEMENT Refine clusters by regrouping 4 – MOTIF FINDING Submit sequences in each cluster to Block Maker compare blocks using LAMA Cluster blocks by using BAG I.U. School of Informatics A Case Study with Disease Resistance Genes in Arabidopsis thaliana I.U. School of Informatics Why Disease Resistance Genes? I.U. School of Informatics Background, Disease Resistance Genes Domain Probable Function TIR CC KIN LRR Recognition of specificity NB ATP and GTP binding I.U. School of Informatics Case Study, Arabidopsis thaliana • 116 disease resistance protein or disease resistance protein like annotated sequences were extracted from Arabidopsis thaliana genome • Clustered into 32 groups • 20 to 640 sequences were added in each cluster after HMM iterations • After refinement four clusters were formed for further analysis # of Sequences Cluster 1 96 Cluster 2 45 Cluster 3 641 Cluster 4 11 I.U. School of Informatics Case Study, Arabidopsis thaliana PFAM Search Domains Cluster 1 NB-ARC, TIR, Kin, LRR Cluster 2 NB-ARC, Kin, LRR Cluster 3 Ser/Thr Kin Cluster 4 Kin I.U. School of Informatics Case Study, Arabidopsis thaliana Results, Block Maker 15218608 15218618 15220795 Cluster1 Cluster2 I.U. School of Informatics YDVFLSFRGVDTRQTIVSHL YDVFLSFRGEDTRKNIVSHL YDVFLSFRGEDTRKTIVSHL Case Study, Arabidopsis thaliana Clusters at the whole gene level Results, Lama and BAG Cluster1 Cluster2 Cluster1 Cluster2 Cluster3 Clusters at the Block Level I.U. School of Informatics Clusters at the whole gene level Case Study, Arabidopsis thaliana RPS4 RPP1 RPP5 Cluster1 TIR-I Cluster2 TIR-II Kin1a Kin1a NBS-A Kin2 NBS-B Kin2 NBS-B NBS-C GLPL RPP8 RPM1 Cluster1 Cluster2 Cluster3 Clusters at the Block Level I.U. School of Informatics LRR LRR Case Study, Arabidopsis thaliana Number of Disease Resistance Gene Candidates on each Chromosome CHR-1 CHR-II CHR-III CHR-IV CHR-V Cluster 1 Cluster 2 I.U. School of Informatics 16 20 2 0 6 6 16 4 35 9 Case Study, Arabidopsis thaliana New Disease Resistance Gene Candidates Cluster 1 Cluster 2 GI 15236505 GI 15242136 GI 15233862 GI 15221277 GI 15221280 GI 15217940 GI 15221744 I.U. School of Informatics Case Study, Arabidopsis thaliana To test effectiveness of the computational procedure 792 Unique sequences were merged and submitted to MEME and PRATT to detect functional motifs. • Time : Took more than 9000 minutes on Pentium IV 1.7 GHz machine running on Linux • Result : No known disease resistance gene motifs were detected I.U. School of Informatics Case Study, Arabidopsis thaliana CONCLUSIONS: Sensible combination of tools provides an excellent mechanism for motif detection Clustering helps to improve performance of other well known tools I.U. School of Informatics ACKNOWLEDGEMENT Motif Discovery from Large Number of Sequences: A Case Study with Disease Resistance Genes in Arabidopsis thaliana Irfan Gunduz, Sihui Zhao, Mehmet Dalkilic and Sun Kim will be presented at The 2003 International Conference on Mathematics and Engineering Techniques in Medicine and Biological Sciences I.U. School of Informatics Case Study, Arabidopsis thaliana I.U. School of Informatics Disease Resistance Mechanism I.U. School of Informatics COMPUTATIONAL PROCEDURE Refinement B A D C I.U. School of Informatics B D C