(精选)Bayes一教学课件.pptVIP

Bayesian Learning of MicroRNA Targets from Sequence and Expression Data Author: Jim C. Huang etc. Lecturer: Dong Yue Director: Dr. Yufei Huang Outline The background of miRNA. The biology model A Bayesian model for mRNA regulation Algorithm evaluation Validating GenMiR++-predicted let-7b targets What is miRNA? MicroRNAs (miRNA) are single-stranded RNA molecules of about 21-23 nucleotides in length thought to regulate the expression of other genes. The Process of miRNA A movie to explain miRNA. /?CategoryID=174 Biology Problem Biologist can not do experiment for all mRNAs and miRNAs . Some many targets been predicted by TargetScan which is a very popular miRNA target algorithm. TargetScan use sequence level data, can we use so other kind of data? Microarray data Red(solid) curve means the expression of mRNA which was repressed by miRNA in specific tissue. Blue(dashed) curve means the background distribution which is the normal expression mRNA. Thinking the Model biologically for only one target Looking at the miRNA target which is predicted by TargetScan. What is TargetScan. 2 If this miRNA is highly expressed in a given tissue? 3. Whether the expression of a targeted transcript is negatively shifted with respect to a background expression level. If 2,3 is Yes, it is very likely a target in reality. The goal of Bayesian Model Find the posterior probability: Graphical model A Bayesian model for mRNA regulation Why we need an approximate method Require integrating over the parameters Sum over an exponential number of combinations of miRNA interactions per mRNA. Variational Bayesian Learning of miRNA targets Observed variables v: X, Z, and C Unobserved variables u: S Model parameters η:Λ Γ The exact posterior is: The sorrogate distribution is: KL-divergence: Variational Bayesian Learning of miRNA targets Define: represents the probability that miRNA k targets mRNA g given the data. represents the expected values of the