Transcript lecture15x
CS173 Lecture 15: TF Motifs (Harendra) MW 11:00-12:15 in Beckman B302 Prof: Gill Bejerano TAs: Jim Notwell & Harendra Guturu http://cs173.stanford.edu [BejeranoWinter12/13] 1 Announcements • Project milestones due Today http://cs173.stanford.edu [BejeranoWinter12/13] 2 Review: Transcriptional regulation of genes Enhancer (CRM) Transcription Start Site (TSS) Thousands of transcription factor-CRM interactions that control gene expression in each cell type http://cs173.stanford.edu [BejeranoWinter12/13] 3 Last Time: ChIP-Seq - a first glimpses of the regulatory genome in action Peak Calling Cis-regulatory peak http://cs173.stanford.edu [BejeranoWinter12/13] 4 Last Time: Infer functions of ChIP-seq binding profile using GREAT Gene transcription start site SRF binding ChIP-seq peak Ontology term (e.g. ‘actin cytoskeleton’) GREAT = Genomic Regions Enrichment of Annotations Tool p = 0.33 of genome annotated with n = 6 genomic regions k = 5 genomic regions hit annotation P = Prbinom(k ≥5 | n=6, p =0.33) http://cs173.stanford.edu [BejeranoWinter12/13] 5 GREAT gives you a tables of functions Top GREAT enrichments of SRF Ontology Term # Genes Binomial Experimental P-value support* 30 31 7x10-9 5x10-5 Miano et al. 2007 Pathway Commons TRAIL signaling 32 Class I PI3K signaling 26 5x10-7 2x10-6 Bertolotto et al. 2000 TreeFam FOS gene family 5 1x10-8 Chai & Tarnawski 2002 TF Targets Targets of SRF Targets of GABP Targets of YY1 Targets of EGR1 84 28 44 23 5x10-76 4x10-9 1x10-6 2x10-4 Positive control Gene Ontology actin cytoskeleton actin binding Miano et al. 2007 Poser et al. 2000 ChIp-Seq support Natesan & Gilman 1995 * Known from literature – as in function is known, SOME of the genes are known, and the binding sites highlighted are NOT. http://cs173.stanford.edu [BejeranoWinter12/13] 6 Last Time: Infer functions of ChIP-seq binding profile using GREAT Gene transcription start site SRF binding ChIP-seq peak π Ontology term (e.g. ‘actin binding’) π π π GREAT = Genomic Regions Enrichment of Annotations Tool p = 0.5 of genome annotated with π n = 6 genomic regions kπ = 4 genomic regions hit annotation P = Prbinom(kπ≥4 | n=6, pπ =0.5) http://cs173.stanford.edu [BejeranoWinter12/13] 7 GREAT gives you a tables of functions Top GREAT enrichments of SRF Ontology Term # Genes Binomial Experimental P-value support* 30 31 7x10-9 5x10-5 Miano et al. 2007 Pathway Commons TRAIL signaling 32 Class I PI3K signaling 26 5x10-7 2x10-6 Bertolotto et al. 2000 TreeFam FOS gene family 5 1x10-8 Chai & Tarnawski 2002 TF Targets Targets of SRF Targets of GABP Targets of YY1 Targets of EGR1 84 28 44 23 5x10-76 4x10-9 1x10-6 2x10-4 Positive control Gene Ontology actin cytoskeleton actin binding Miano et al. 2007 Poser et al. 2000 ChIp-Seq support Natesan & Gilman 1995 * Known from literature – as in function is known, SOME of the genes are known, and the binding sites highlighted are NOT. http://cs173.stanford.edu [BejeranoWinter12/13] 8 GREAT gives you a tables of functions Top GREAT enrichments of SRF Ontology Term 30 31 7x10-9 5x10-5 Miano et al. 2007 Pathway Commons TRAIL signaling 32 Class I PI3K signaling 26 5x10-7 2x10-6 Bertolotto et al. 2000 TreeFam FOS gene family 5 1x10-8 Chai & Tarnawski 2002 TF Targets Targets of SRF Targets of GABP Targets of YY1 Targets of EGR1 84 28 44 23 5x10-76 4x10-9 1x10-6 2x10-4 Positive control Gene Ontology actin cytoskeleton actin binding Different # Genes Binomial Experimental P-value support* Miano et al. 2007 Poser et al. 2000 ChIp-Seq support Natesan & Gilman 1995 * Known from literature – as in function is known, SOME of the genes are known, and the binding sites highlighted are NOT. http://cs173.stanford.edu [BejeranoWinter12/13] 9 But doing the experiment is the hard part! • Hard or impossible to get the required cells • Some cells don’t occur in enough quantity to ChIP • Others are hard to dissect • Certain human tissues are hard to obtain • Hard to get a good antibody • Ex: We have ChIP results for a factor in brain • We have not be able to repeat it since we can’t find the same antibody • Lots of time and money to do one experiment • Only information for one context – cell type or time Can we computationally predict the binding sites for many contexts and factors? http://cs173.stanford.edu [BejeranoWinter12/13] 10 Recall: TFBS Position Weight Matrix (PWM) Experimentally determined sites A T G G C A T G A G G G T G C G A T C G C A T G T T G C C A C G A T G G T A T T A T T C G A C G A G G G C G T T A T G A C A T G A T G G C A T G A C T G G A T G Alignment (count) Matrix A 9 0 0 1 0 8 0 C 0 1 1 1 7 0 3 G 0 2 7 8 1 2 0 T 1 7 2 0 2 0 7 Frequency Weight Matrix A 0.9 0.0 0.0 0.1 0.0 0.8 0.0 C 0.0 0.1 0.1 0.1 0.7 0.0 0.3 G 0.0 0.2 0.7 0.8 0.1 0.2 0 T 0.1 0.7 0.2 0.0 0.2 0.0 0.7 Cons A T G G C A T 0 0 8 2 0.0 0.0 0.8 0.2 G Can we use a PWM to predict where the TF will bind in the genome (without doing ChIP-seq)? http://cs173.stanford.edu [BejeranoWinter12/13] 11 Binding Site Prediction using Match Problem: High number of false positives. http://cs173.stanford.edu [BejeranoWinter12/13] 12 Recall: TFBS Position Weight Matrix (PWM) Experimentally determined sites A T G G C A T G A G G G T G C G A T C G C A T G T T G C C A C G A T G G T A T T A T T C G A C G A G G G C G T T A T G A C A T G A T G G C A T G A C T G G A T G Alignment (count) Matrix A 9 0 0 1 0 8 0 C 0 1 1 1 7 0 3 G 0 2 7 8 1 2 0 T 1 7 2 0 2 0 7 Frequency Weight Matrix A 0.9 0.0 0.0 0.1 0.0 0.8 0.0 C 0.0 0.1 0.1 0.1 0.7 0.0 0.3 G 0.0 0.2 0.7 0.8 0.1 0.2 0 T 0.1 0.7 0.2 0.0 0.2 0.0 0.7 Cons A T G G C A T 0 0 8 2 0.0 0.0 0.8 0.2 G 1.2 0.7 0.7 0.7 0.6 1.0 0.8 1.0 Information content of each column Information content of a motif = sum of all columns = 1.2 + 0.7 + 0.7 +0.6 + 1.0 + 0.8 + 1.0 = 6.0 http://cs173.stanford.edu [BejeranoWinter12/13] 13 Information content is a measure of motif specificity SRF (IC ~ 12) SPIB (IC ~ 5) REST (IC ~ 25) How do these compare to a library of many PWMs? http://cs173.stanford.edu [BejeranoWinter12/13] 14 PWMs have a range of information content SRF SPIB REST http://cs173.stanford.edu [BejeranoWinter12/13] 15 Information content determines how accurately we can predict the binding site • Measure of motif specificity SRF SRF 2 million http://cs173.stanford.edu [BejeranoWinter12/13] 16 Information content determines how accurately we can predict the binding site • Measure of motif specificity SRF 2 million matches to the SRF motif, SRF but ChIP-seq and other estimates suggest ≈ 10,000 actual binding sites 2 million Can we do better? http://cs173.stanford.edu [BejeranoWinter12/13] 17 Use excess conservation to improve prediction accuracy Aaron Shoa Wenger et al., PRISM offers a comprehensive genomic approach to transcription factor function prediction. 2013 http://cs173.stanford.edu [BejeranoWinter12/13] 18 Use shuffled motifs to calculate confidence of excess conservation binding site prediction Transcription factor motif Genome-wide binding site predictions 10 Shuffled Transcription factor motifs Genome-wide binding site predictions fraction conserved Confidence is the fraction conserved in excess. confidence = excess / total real total = 0.32 excess = 0.12 shuffled branch length (subst / site) http://cs173.stanford.edu [BejeranoWinter12/13] 19 Probabilistic interpretation • Confidence is the probability that a motif instance is functional given its observed conservation. PrR(functional | C ≥ c) = 1 - PrR(not functional | C ≥ c) PrR(C ≥ c | not F) PrR(not F) =1PrR(C ≥ c) R: real motif S: average shuffled motif PrR(C ≥ c) PrR(C ≥ 1.5) = 0.2 PrS(C ≥ c) branch length (subst / site) PrS(C ≥ c) PrR(not F) =1PrR(C ≥ c) PrR(C ≥ c) - PrS(C ≥ c) PrR(not F) = PrR(C ≥ c) PrR(C ≥ c) - PrS(C ≥ c) excess ≈ = PrR(C ≥ c) total http://cs173.stanford.edu [BejeranoWinter12/13] 20 Excess conservation score defined by genomic background http://cs173.stanford.edu [BejeranoWinter12/13] 21 Excess conservation score also defined by motif http://cs173.stanford.edu [BejeranoWinter12/13] 22 Perform genome-wide binding site predictions… ARE THE PREDICTIONS ANY GOOD? http://cs173.stanford.edu [BejeranoWinter12/13] 23 Use ChIP-seq overlap as a measure of sensitivity Genome-wide binding site predictions for one factor (Ex: E2F4) ChIP-seq for same factor (Ex: E2F4) Sensitivity = Overlapping ChIP-peaks / Total ChIP-peaks But how do you assess if your overlap is good? Compare to the best tool out there (or all the tools, if there is no “best”) http://cs173.stanford.edu [BejeranoWinter12/13] 24 Excess conservation binding site prediction is more accurate than existing methods (prior state of the art) http://cs173.stanford.edu [BejeranoWinter12/13] 25 conservation (% identity) Excess conservation captures binding site profile similar to ChIP-seq http://cs173.stanford.edu [BejeranoWinter12/13] 26 Submit predictions to GREAT • Now we have good genome-wide binding site predictions for many factors Lets submit them to GREAT and find out what they are doing… http://cs173.stanford.edu [BejeranoWinter12/13] 27 Comparing binding site prediction to ChIP-seq Transcription factor GABPA REST (NRSF) SRF In Jurkat STAT3 In mESC Ontology GO Biological Process GO Cellular Component GO Molecular Function Mouse Phenotypes PANTHER Pathway Pathway Commons GO Biological Process GO Cellular Component GO Molecular Function Mouse Phenotypes PANTHER Pathway Pathway Commons GO Biological Process GO Cellular Component GO Molecular Function Mouse Phenotypes PANTHER Pathway Pathway Commons GO Biological Process GO Molecular Function Mouse Phenotypes Pathway Commons Top-ranked biological context translation membrane coat translation initiation factor activity increased single-positive T cell number general transcription by RNA polymerase I transcription neurotransmitter transport neuronal cell body cation channel activity abnormal synaptic transmission synaptic vesicle trafficking transmission across chemical synapses muscle structure development actin cytoskeleton structural constituent of muscle dilated heart ventricles cytoskeletal regulation by Rho GTPase regulation of insulin secretion by acetylcholine negative regulation of signal transduction transforming growth factor beta binding abnormal spleen B cell follicle morphology Signaling events mediated by TCPTP http://cs173.stanford.edu [BejeranoWinter12/13] GREAT rank for ChIP-seq 2 14 4 None 1 3 1 None 1 1 2 3 None 1 None None None None None None None None Experimental support (Genuario and Perry, 1996) Novel (Genuario and Perry, 1996) (Yu et al., 2010) (Hauck et al., 2002) (Hauck et al., 2002) (Schoenherr et al., 1996) (Schoenherr et al., 1996) (Schoenherr et al., 1996) (Schoenherr et al., 1996) (Schoenherr et al., 1996) (Schoenherr et al., 1996) (Miano et al., 2007) (Miano et al., 2007) (Miano et al., 2007) (Parlakian et al., 2004) (Hill et al., 1995) Novel (Naka et al., 1997) (Kinjyo et al., 2006) (Schmidlin et al., 2009) (Yamamoto et al., 2002) 28 PRISM re-discovers known functions TF SRF function muscle structure development p-value 7.43×10-41 target genes 157 GLI2 skeletal system development 7.07×10-48 192 Is the number of re-discovered known functions impressive? CRX retinal photoreceptor degeneration 1.30×10-10 AR abnormal spermiogenesis 1.19×10-6 http://cs173.stanford.edu [BejeranoWinter12/13] 34 26 29 Evaluate re-discovery of known function using “closed loops” How can we assess if the functional associations predicted by PRISM for a particular TF are reasonable without reading a lot of papers? One way is to check if the TFs are annotated with the function (form a closed loop) Genes involved in “muscle structure development” SRF Is SRF itself annotated with the term “muscle structure development”? SRF YES – a “closed loop” http://cs173.stanford.edu [BejeranoWinter12/13] 30 PRISM predictions are consistent with known transcription factor biology Null Model: How many closed loops using 50,000 random shuffled PWM libraries? http://cs173.stanford.edu [BejeranoWinter12/13] 31 Many non-closed loops are still true TF function GATA6 abnormal pancreas development SRF actin cytoskeleton p-value 5.69×10-13 4.84×10-58 target genes 23 142 1. Incomplete annotation Nature Genetics, December 2011. 2. “Regulation of” annotation SRF acts in the nucleus, where it regulates actin cytoskeleton genes. http://cs173.stanford.edu [BejeranoWinter12/13] 32 Raw GREAT results need cleaning for conserved TFBS • Now we have good genome-wide binding site predictions for many factors • AND we have functional predictions without ChIP-seq Was it as easy as creating binding sites and submitting the results to GREAT? …not quite… http://cs173.stanford.edu [BejeranoWinter12/13] 33 Shuffled motifs also give GREAT enrichments Transcription factor motif Genome-wide Run GREAT binding site and observe predictions biological function 10 Shuffled Transcription factor motifs Genome-wide Run GREAT binding site and observe predictions biological function Filter PRISM Examine closely http://cs173.stanford.edu [BejeranoWinter12/13] 34 Shuffled motifs are used to create a “E-value” metric to black list enrichments that show up for shuffles (from shuffles) Stage 1: GREAT on binding site prediction Obtained = GREAT # TF-term associations TF-term FDR closed loop % 31,946 50.5% 3.3% Stage 2: Top Stage 3: PRISM significant terms (via black GREAT terms listing) Kept Kept = PRISM PRISM vs. GREAT on b.s. prediction 7,529 49.5% 5.3% 1,658 GREAT predictions kept 16.4% FDR improvement 10.9% fraction loops improvement 5.2% 308% 329% What are all the terms we are throwing away? http://cs173.stanford.edu [BejeranoWinter12/13] 35 GREAT enrichments from shuffles are due to conservation bias • Create 10,000 random sets of random conserved non-coding regions • Run GREAT • How do the enrichments compared to those from shuffled motifs? Shuffles (2488) 755 1733 546 CNEs (2279) Pro: E-value helps us get more accurate predictions by removing false predictions Con: Conservation bias filter, causes us to lose potentially real enrichments in systems that are more often conserved http://cs173.stanford.edu [BejeranoWinter12/13] 36 So far… • “Excess Conservation” • advanced the state of the art for binding site prediction • “PRISM pipeline” • combined accurate binding site prediction with GREAT • Publically offered as a web application • bejerano.stanford.edu/prism http://cs173.stanford.edu [BejeranoWinter12/13] 37 The rest of the talk includes pre-publication work http://cs173.stanford.edu [BejeranoWinter12/13] 38