Genome 559 Wi RNA Function, Search, Discovery

Transcription

1 Genome 559 Wi 2009 RN Function, Search, Discovery

2 The Message Cells make lots of RN noncoding RN Functionally important, functionally diverse Structurally complex New tools required alignment, discovery, search, scoring, etc. 2

3 RN Secondary Structure: RN makes helices too Base pairs U C G U G 5 UC C G C G G C U C G G U C C U 3 Usually single stranded 3

4 Fig. 2. The arrows show the situation as it seemed in Solid arrows represent probable transfers, dotted arrows possible transfers. The absent arrows (compare Fig. 1) represent the impossible transfers postulated by the central dogma. They are the three possible arrows starting from protein.

5 Proteins catalyze & regulate biochemistry The Met Repressor SM 5

6 lberts, et al, 3e. The protein way Riboswitch alternative SM Grundy & Henkin, Mol. Microbiol 1998 Epshtein, et al., PNS 2003 Winkler et al., Nat. Struct. Biol

7 lberts, et al, 3e. The protein way Riboswitch alternatives SM-II SM-I Grundy, Epshtein, Winkler et al., 1998, 2003 Corbino et al., Genome Biol

8 lberts, et al, 3e. The protein way Riboswitch alternatives SM-III SM-I SM-II Fuchs et al., NSMB 2006 Grundy, Epshtein, Winkler et al., 1998, 2003 Corbino et al., Genome Biol

9 lberts, et al, 3e. The protein way Riboswitch alternatives SM-III SM-I SM-II SM-IV Grundy, Epshtein, Winkler et al., 1998, 2003 Corbino et al., Genome Biol Fuchs et al., NSMB 2006 Weinberg et al., RN

10 boxed = confirmed riboswitch (+2 more) Widespread, deeply conserved, structurally sophisticated, functionally diverse, biologically important uses for ncrn throughout prokaryotic world. Weinberg, et al. Nucl. cids Res., July :

11 Why is RN hard to deal with? Similar structures, but only 29% seq id G CG U C U G U CU C U G C G U G C C G GCGG G U G G G G G C CU G G GC C C G C GG G G G G C C C U C G U C U C G C G G C G C C U U G G G C G G C U U U CU U G U UUC U CUG C C C G G C G G G G G C U G C CG G U UG U U U C CG U U G U GU G CG 11 : Structure often more important than sequence

12 Motif Description & Inference 12

13 RN Motif Models Covariance Models (Eddy & Durbin 1994) aka profile stochastic context-free grammars aka hidden Markov models on steroids Model position-specific nucleotide preferences and base-pair preferences Pro: accurate Con: model building hard, search sloooow 13

14 14

15 mrn leader mrn leader switch? 15

16 Mutual Information M ij = f f xi,xj xi,xj log xi,xj 2 f xi f xj ; 0 M ij 2 Max when no seq conservation but perfect pairing; Expected score gain from modeling i & j as paired. Given columns, finding optimal pairing without pseudoknots can be done by dynamic programming 16

17 17

18 Fast Motif Search Faster Genome nnotation of Non-coding RNs Without Loss of ccuracy Weinberg & Ruzzo Recomb 04, ISMB 04, Bioinformatics 06 18

19 CM s are good, but slow Rfam Reality EMBL Our Work EMBL Rfam Goal EMBL BLST Ravenna junk CM hits 1 month, 1000 computers CM hits ~2 months, 1000 computers CM junk hits 10 years, computers

20 Name Results: New ncrn s? # found BLST + CM # found rigorous filter + CM # new Pyrococcus snorn Iron response element Histone 3 element Purine riboswitch Retron msr Hammerhead I Hammerhead III U4 snrn S-box U6 snrn U5 snrn U7 snrn

21 Motif Discovery In Prokaryotes (Vertebrates too, but no time today see, e.g., Torarinsson, et al. Genome Research, Jan 2008) 21

22 pipeline for RN motif genome scans CDD CMfinder Ortholgous genes Top datasets Motifs Search Genome database Upstream sequences Footprinter Rank datasets Homologs Yao, Barrick, Weinberg, Neph, Breaker, Tompa and Ruzzo. Computational Pipeline for High Throughput Discovery of cis-regulatory Noncoding RN in 22 Prokaryotes. PLoS Computational Biology. 3(7): e126, July 6, 2007.

23 Identify CDD group members 2946 CDD groups < 10 CPU days Retrieve upstream sequences Footprinter ranking < 10 CPU days CMfinder motifs 1 ~ 2 CPU months Motif postprocessing 1740 motifs RaveNn 10 CPU months CMfinder refinement < 1 CPU month Motif postprocessing 1466 motifs 23

24 boxed = confirmed riboswitch (+2 more) Weinberg, et al. Nucl. cids Res., July :

25 Summary ncrn - apparently widespread, much interest Covariance Models - powerful but expensive RaveNn filtering - search ~100x faster with no/little loss CMfinder - CM-based motif discovery in unaligned sequences Pipelines integrating comp and bio for ncrn discovery Many vertebrate ncrns? structural, not seq conservation; functional significance unclear BIG CPU demands Still need for further methods development & application

26 Course Wrap Up Modern biology is suddenly very data-rich Mathematical & computational tools needed We showed: sequence modeling, alignment & search, phylogeny, linkage/association mapping, some data bases Python is a good tool for doing much of this There s lots more! Check out, e.g., GENOME 540/1, CSE

27 We hope you enjoyed it. Thanks! 27