Searching for Noncoding RNA

Transcription

1 Searching for Noncoding RN Larry Ruzzo omputer Science & Engineering enome Sciences niversity of Washington Bio 2006, Seattle, 8/4/2006 1

2 Outline Noncoding RN Why are they hard to discover? Key computational problems: Motif discovery Motif search Sketch new methods pplication: cis-regulatory motifs in actinobacteria 2

3 Non-coding RN Messenger RN - codes for proteins Non-coding RN - all the rest Before, say, mid 1990 s, 1-2 dozen known (critically important, but narrow roles: e.g. trn) Since mid 90 s dramatic discoveries Hundreds of new families Regulation, transport, stability/degradation E.g. microrn : 100 s in humans By some estimates, ncrn >> mrn 3

4 Example: lycine Regulation How is glycine level regulated? Plausible answer: g TF g transcription factors (proteins) g TF g gce protein glycine cleavage enzyme gene g DN 4

5 The lycine Riboswitch ctual answer (in many bacteria): g g gce protein g g 5 3 gce mrn glycine cleavage enzyme gene DN Mandal et al. Science

6 6S mimics an open promoter E.coli Barrick et al. RN 2005 Trotochaud et al. NSMB 2005 Willkomm et al. NR

7 7 Why should these be hard to discover? : Structure often more important than sequence

8 Wanted ood, fast search tools ( RN BLST, etc.) ood, fast motif discovery tools ( RN MEME, etc.) Importance of structure makes both hard; progress on both below 8

9 How to model an RN Motif? onceptually, start with a profile HMM: from a multiple alignment, estimate nucleotide/ insert/delete preferences for each position given a new seq, estimate likelihood that it could be generated by the model, & align it to the model mostly del ins all 9

10 How to model an RN Motif? dd column pairs and pair emission probabilities for base-paired regions <<<<<<< paired columns >>>>>>> 10

11 RN Motif Models ovariance 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 on: model building hard, search sloooow 11

12 Task 1: Faster Search 12

13 M s are good, but slow Rfam Reality EMBL Our Work EMBL Rfam oal EMBL BLST Ravenna M M M junk hits 1 month, 1000 computers hits ~2 months, 1000 computers junk hits 10 years, 1000 computers 13

14 Ravenna: enome Scale RN Search Typically 100x speedup over raw M, with no (or little) loss in accuracy: drop structure from M to create a (faster) HMM use that to pre-filter sequence; discard parts where, provably, the M will score < threshold; actually run M on the rest (the promising parts) assignment of HMM transition/emission scores is key (large convex optimization problem) Weinberg & Ruzzo, Bioinformatics, 2004,

15 Results: buried treasures Name # found BLST + M # found rigorous filter + M # new Pyrococcus snorn Iron response element Histone 3 element Purine riboswitch Retron msr Hammerhead I Hammerhead III snrn S-box snrn snrn snrn

16 Task 2: Motif Discovery 16

17 RN Motif Discovery Typical problem: given a ~10-20 unaligned sequences of ~1kb, most of which contain instances of one RN motif of, say, 150bp -- find it Example: 5 TRs of orthologous glycine cleavage genes from γ-proteobacteria 17

18 Obvious pproach I Predict secondary RN structure using MFOLD or Vienna Problems false folding predictions comparing structures 18

19 Obvious pproach II: Predict from Multiple Sequence lignment ompensatory mutations reveal structure, but usual alignment algorithms penalize them (twice) 19

20 Our pproach: Mfinder Simultaneous alignment, folding and M-based motif description using an EM-style learning procedure Yao, Weinberg & Ruzzo, Bioinformatics,

21 lignment M lignment Similar to HMM, but much slower Largely from Eddy & Durbin, 94 But new way to infer which columns to pair, via a principled combination of mutual information and predicted folding energy 21

22 MFinder Harder: Finding Ms without alignment Yao, Weinberg & Ruzzo, Bioinformatics, 2006 Folding predictions Smart heuristics andidate alignment M Search Realign EM 22

23 Mfinder ccuracy (on Rfam families with flanking sequence) /W /W 23

24 Task 3: pplication enome-wide search for cis-regulatory RN elements (in prokaryotes, initially) 24

25 oal: Predicting New cis-regulatory RN Elements iven unaligned TRs of coexpressed or orthologous genes, find common structural motifs Difficulties: Low sequence similarity: alignment difficult Varying flanking sequence Motif missing from some input genes 25

26 pproach hoose a bacterial genome For each gene, get close orthologs (DD) Find most promising genes, based on conserved sequence motifs (Footprinter) From those, find structural motifs (Mfinder) enome-wide search for more instances (Ravenna) Expert analyses (Breaker Lab, Yale) 26

27 enome Scale Search: Why Most riboswitches, e.g., are present in ~5 copies per genome Throughout (most of) clade More examples give better model, hence even more examples, fewer errors More examples give more clues to function - critical for wet lab verification 27

28 enome Scale Search Mfinder is directly usable for/with search Folding predictions Smart heuristics andidate alignment M Search Realign 28

29 Results Process largely complete in bacillus/clostridia gamma proteobacteria cyanobacteria actinobacteria nalysis ongoing 29

30 Some Preliminary ctino Results 8 of 10 Rfam families found Rfam Family Type (metabolite) Rank THI riboswitch (thiamine) 4 ydao-yua riboswitch (unknown) 19 obalamin riboswitch (cobalamin) 21 SRP_bact gene 28 RFN riboswitch (FMN) 39 yybp-ykoy riboswitch (unknown) 48 gcvt riboswitch (glycine) 53 S_box riboswitch (SM) 401 tmrn gene Not found RNaseP gene Not found not cisregulatory (got one anyway) 30

31 More Prelim ctino Results Many others (not in Rfam) are likely real of top 50: known (Rfam, 23S) 10 probable (Tbox, IRE, Lex, parp, pyrr) 7 probable (ribosomal genes) 9 potentially interesting 12 unknown or poor 12 One bench-verified, 2 more in progress 31

32 mrn leader mrn leader switch? 32

33 Ongoing & Future Work Still automating a few steps, e.g. identifying duplicates Improved ranking/motif significance stats Better ortholog clustering Performance & scale-up Eukaryotic mrns, e.g. TRs 33

34 cknowledgements Zizhen Yao (W) Zasha Weinberg (W Yale) Shane Neph (W) Martin Tompa (W) Ron Breaker (Yale) Jeff Barrick (Yale) 34