Motif finding

CREATED: 200701021353 ** Motivation

• orthologous promoted regions

** Challenges

• transcription start site vs traslation start site

** Methods

• pattern driven - search 4^l patterns
• sample dirver
• extended sample driven approach ** check k neighbour of l-mers in sample ** number of patterns of each l-mer = (l choose k)3^k
• pattern branching approach ** best neighbours
• PWM driven ** partition the space of motifs
• multiple alignment
• cluster motifs + TRANSFAC

** Ranking Motifs

• comparing motifs with different lengths using E-value = size of search space x p-value (Keich)

** Models

• (l,d) motif is not realisticas hamming distance does not take into account fact that certain positions are conserved
• PWM is more accurate