Techniques for estimating local alignment score distribution parameters

[cb1025]

To score a sequence alignment, the distribution of local alignment scores involving random sequences is used as a statistical reference.

What are some current techniques used by computational biologists in estimating the parameters of such distributions? How does this distribution vary with different species? Or maybe I should ask: how do we take into account of the differences between species when modeling these distributions?

I recently came across the "island method" approach but I don't quite understand it. > < Could someone please offer a short explanation?

Are there example implementations or opensource codes of these parameter estimation methods I could check out?

Thank you very much

[gtuckerkellogg]

I think you are referring to the island method described in Altschul, S.F., Bundschuh, R., Olsen, R. & Hwa, T. The estimation of statistical parameters for local alignment score distributions. Nucleic Acids Res. 29, 351–361 (2001). I've seen it used in other papers, such as the original miRDeep reference, but haven't seen the implementation (as code) in any paper.

[cb1025]

Yes, thank you! MiRDeep used it to estimate parameters for the Gumbel distribution. However, I have not been able to find a full, step-by-step example, let along one with implementation details.

I have also been looking at some of Poleksic's work such as: Island method for estimating the statistical significance of profile-profile alignment scores. BMC Bioinformatics. 2009 Apr 20;10:112. doi: 10.1186/1471-2105-10-112.

I feel I also need to read up on some the basic statistics behind it (e.g. Gumbel EJ: Statistics of Extremes. Columbia University Press, New York, NY; 1958.) to really understand what's going on.

Please do advise on a good approach for a starter like me! Thanks again