Computational Motif Finding

Computational Motif Finding

We can use microarray or SAGE detects the expression of every gene at a certain cell state. We can also use clustering to find genes that are co-expressed (potentially share regulation). To decode gene regulation, look upstream regions of genes always expressed together. Proteins conbind to the DNA by recognizing specific DNA sequence. Some will turn on the gene. Some will turn off the gene. Some control the amount of gene expression. These proteins are called TFs. They can regulate the expression of nearby gene. TFs bind motifs. Motifs can be computational discovered when there are enough cases for machine learning.

Input data: 

Upstream sequences of gene expression profile cluster.

20-800 sequences, each 300-5000 bps long.

Output: 

Enriched sequence patterns (motifs)

Ultimate goals: 

1. Which TFs are involved and their binding motifs and effects (activate/enhance/repress gene expresssion)？

2. Which genes are regulated by this TF, why is there disease when a TF goes wrong?

3. Are there binding partner / competitor for a TF?

Challenges: 

1. Where/what the signal.The motif should be abundant and abundant with significance.

2. Double stranded. DNAMotif appears in both strands

3. Base substitutions.Proteins allow mistakes. TF is the protein that recognize sequence which is kind of a structure, as roughly, when the structure are right even there are some mismatches here and there. Sequences do not have to match the motif perfectly, base substitutions are allowed.

4. Variable motif copies.Some sequences do not have the motif. Some have multiple copies of the motif. A group of gene are co-expressed, but they may not be co-regulated. Co-expression doesn't mean co-regulation.

TF Motif Representation: 

1. Sometimes we use degenerate motif. (IUPAC)

2. Position weight matrix (PWM): need score cutoffSome sequence in front of the gene (whether it is the motif or not). It is not so clear, because every sequence will be able to get a score.

Segment ATGCAGCT score = p(generate ATGCAGCT from motif matrix)/p(generate ATGCAGCT from background)

A Word on Sequence Logo: 

SeqLogo consists of stacks of symbols, one stack for each position in the sequence.The overall height of the stack indicates the sequence conservation at that position.The height of symbols within the stack indicates the relative frequency of nucleic acid at that position

TF Motif Database:

--- JASPAR: literature curation

--- UniProbe: Protein Binding Microarrays ( very popular,Badis et al, Science 2009)

--- Factor Book and HOMER:--ChIP-seq