Monday, June 18, 2007

calculate PWM similarity

Overview:
Most methods calculate the similarity of each column, and then sum all columns up.

For each column, people have used the following measures:
1. Pearson correlation coeffecient [8871566, 10698627, 15735639]
2. K-L distance [15980506 (with web tool), 12015892, 14534164]
3. Euclidean distance [14985506]
4. Kai-square test [15319260 (dedicated paper)]
5. Fisher's Exact test [15319260]
6. Average log likelihood ratio [14668220]
7. SW [15066426]

Shobhit Gupta et al. come up with a idea to use p-value measure column similarity[17324271]. The calculation of p-value can be based on any of the above seven measures.