MLOGD: Notes

Notes on Monte Carlo simulations:

MLOGD provides an estimate of the probability ratio for the null model versus the alternate model. However, it doesn't directly provide confidence limits on this estimate. One way to estimate confidence limits is to calculate the distribution of scores observed for a large number of simulated alignments with the same initial reference sequence and pairwise divergences as the input alignment. The simultations involve applying random mutations to the reference sequence, using the nucleotide, codon and amino acid substitution matrices, together with either the null or alternate model CDS annotation, to weight the mutation probabilities at each nucleotide.

On the Monte Carlo simulations page, you can

Run simulations for each reference - non-reference sequence pair using either the null model or the alternate model. Typically you'd run about 50-100 simulations for each model, using the same pairwise divergence as the original reference - non-reference sequence pair. The width of the distribution of the MLOGD scores for the simulated sequence pairs can be used to estimate confidence limits on the MLOGD score for the original reference - non-reference sequence pair.
Run simulations for a range of pairwise divergences using either the null model or the alternate model. The MLOGD scores for the reference - simulated sequence pairs can be used to plot the general distribution of the MLOGD score as a function of pairwise sequence divergence for the particular reference sequence, CDS annotation and query ORF.

In addition, on the Monte Carlo simulations results page, you will get plots for the N123 (i.e. 1st/2nd/3rd codon position mutation fraction) and NsNn (i.e. synonymous/nonsynonymous mutation fraction) statistics introduced in Firth A. E., Brown C. M., 2005, Detecting overlapping coding sequences with pairwise alignments, Bioinformatics, 21, 282-92, together with null model versus alternate model likelihood ratios calculated by comparing the observed statistics with the distribution of null and alternate model scores obtained from the simulations.