RNA Folding

RNA Folding Algorithms Intuitively: given a sequence, find the structure with the maximal number of base pairs For nested structures, four possibilities for S(i,...,j) i,j are paired, added to S(i+1,...,j-1) i is unpaired, added to S(i+1,...,j) j is unpaired, added so S(i,...,j-1) i,j are paired but not to each other, to S(i,...,k), S(k+1,...,j)

RNA Folding by DP Fill in a matrix of S(0,...,seq_length)

RNA Folding Assumptions RNA folding algorithms typically detect only nested structures and do not recognize pseudoknots Some folding algorithms identify pseudoknots but they are typically inefficient or limited (e.g., do not take stacking-dependent pairing models) Current algorithms get about 50-70% of the base pairs correct, on average

MicroRNA Identification

miRNAs are genomically encoded small RNAs processed into single stranded mers incorporated into RNP complex (miRISC) miRISC binds to 3’UTRs, repression of translation modest mRNA degradation MicroRNAs: Introduction miRISC Ago1 Bartel, Cell 116, 2004

MicroRNA Transcription miRNA genes can be in intergenic and intronic regions miRNA genes can be clustered and co-expressed Estimates: 60% singletons, 25% introns, 15% clusters

MicroRNA Examples

MicroRNA Gene Conservation Some miRNAs are highly conserved (e.g. let-7) Conservation must preserve a dsRNA hairpin from which the miRNA is processed by Dicer

MicroRNA Gene Identification MicroRNA Cloning Map cloned ~22nt small RNAs to the genome Predict pre-miRNA secondary structures using m-fold Score pre-miRNAs based on known miRNA precursors Computational Identification Identify conserved genomic segments Predict pre-miRNA secondary structures using m-fold Scoring pre-miRNAs based on the known miRNA precursors

MirScan, MirSeeker, … MicroRNA Gene Identification More complex methods: additional features

MiRBase ~4500 miRNAs in 41 eukaryotes Examples: 474 human, 78 fly Eight viruses express microRNAs

MiRBase

MicroRNAs: Open Questions Promoter Transcritpional start site Transcriptional Termination Transcriptional complex Regulation of miRNA expression

MicroRNA Targets: Mechanism & Identification

Are All RNAs Regulated by miRNAs?

The Target Prediction Problem Target sites show imperfect sequence complementarity: Strong match in 5’ region (‘seed’) Varying complementarity on 3’ end Computational target predictions: Sensitive to exact pairing rules ~100 targets per miRNA within fly transcriptome ~25% of transcriptome under miRNA regulation 3’5’ mRNA 3’5’ miRNA seed Existing algorithms focus on quality of the sequence match between miRNA and mRNA target introduce various filters, e.g. evolutionary conservation 3’ 5’ mRNA 3’5’ miRNA Brennecke et al. 05 wtwt seed

miRanda Target prediction: sequence-based rules miRNA-target complementarity (strong in 5’, weaker in 3’) Refinement with binding free energy scores Use conservation to increase signal to noise

PicTAR: Combinatorial Targets mRNA Perfect nucleus Imperfect nucleus miRNA Filter - over 33% of mature miRNA binding energy to perfect complementary site

PicTAR: Combinatorial Targets Anchor

PicTAR: Combinatorial Targets

Prior (transition) probabilities p0p0 p1p1 p2p2 p3p3 pmpm... Emission probabilities A C U G ACUGUAC GGCAUUAC Generated mRNA U ACUGUAC C GGCAUUAC ACUGCAC... - Independency of binding sites (no overlapping) - Transition does not depend on current state (memoryless) - Competition between background and miRNA 1…m miRNAs Hidden states b

Accessibility: The Missing Component What about target accessibility? miRISC vs.

Experimental Method Drosophila tissue culture cells (S2) No miRNA overexpression establish miRNA expression profile use endogenous miRNA ( copies per cell) (bantam, miR-2 family, miR-184) Dual luciferase reporter assay mutate target site sequence mutate sequence surrounding the target site to alter mRNA secondary structure firefly 3’UTR Renilla UTR engineering Renilla experiment, firefly as internal control mild overexpression of target sequence (<10fold) no target degradation (20h transfection) sensitive, quantitative, linear assay

3’UTR AAAAA target site ~200 b N: ~200 bp fragment, native structure C: ~200 bp fragment, closed structure ’UTRNCC3C3+C5C5+ normalized luciferase ratio target site 5’ end A GA 5 CUCAUCAAAGC UUGUGAUA 3’ 3’ GAGUAGUUUCG GACACUAU 5’ C ACC rpr (miR-2) Target miRNA The Role of Secondary Structure

Target Accessibility Matters ’UTRNCC3C3+C5C5+ normalized luciferase ratio target site 5’ end A GA 5 CUCAUCAAAGC UUGUGAUA 3’ 3’ GAGUAGUUUCG GACACUAU 5’ C ACC rpr (miR-2) Target miRNA grim (miR-2) 3’UTR NC A GCA U GCUC AUCAAAGC UUGUGAU CGAG UAGUUUCG GACACUA ACC U C AAUUAGUUUUCA AAUGAUCUCG UUAGUCGAAAGU UUACUAGAGU U hid (bantam) 3’UTR NC

Accessibility as Important as Sequence ’UTRNCC3C3+C5C5+ normalized luciferase ratio target site 5’ end A GA 5 CUCAUCAAAGC UUGUGAUA 3’ 3’ GAGUAGUUUCG GACACUAU 5’ C ACC rpr (miR-2) Target miRNA A GA CUCAUCAAAGC UUGUGAUA ACC D5D5 D 5+3 G M2M3M6I5 0.7 D5D5 D 5+3 target site mutations

Thermodynamic miRNA::RNA Model

UTR ∆G = ∆G 5 = -15.1∆G 3 = Thermodynamic miRNA::RNA Model

UTRCDSPoly(A) ∆G 0 = ∆G 1 = ∆G open = ∆G 0 - ∆G 1 folding area = target +70bp Thermodynamic miRNA::RNA Model

normalized luciferase ratio D G duplex DD G grim hid rpr 22 constructs altering accessibility of target sites in rpr, hid, grim r=0.36 p<0.11 r=0.7 p<4x DD G with flank 17 up, 13 down r=0.77 p<3x r exploring flank size downstream (bp) upstream (bp) ddG Predicts Measured Repression

ddG differential measured repression differential miR-184 targets r= validated targets 3’ 5’ mRNA 3’5’ miRNA seed Native Target Analysis 12 miR-184 targets with weaker 3’ pairing, tested in different backgrounds to alter secondary structure non-redundant set of 190 experimentally tested miRNA:mRNA target pairs in Drosophila

miRNA target seeds favor highly accessible regions of the genome D G open overrepresentation vs. random accessibility ( D G open ) accessibility ( D G open ) fly human Genome-Wide Target Analysis

Assignment Download the set of human microRNAs Download the set of human UTRs Download the mFold software For each microRNA, identify the set of targets on each UTR, defined by a perfect match to the microRNA seed, bases 2-8 Partition the targets of each microRNA into conserved and non- conserved targets (define a conservation cutoff) Compare the RNA-accessibility of conserved and non- conserved targets for each microRNA For each putative target, extract the 100 bases that surround it Use mFold to compute the free energy of these 100 bases Create a dot-plot with points being microRNAs, and axes being the median (plot #1) or mean (plot #2) free energy of all conserved (x- axis) or non-conserved (y-axis) targets of the microRNA