This track displays the conservation between the human and mouse genomes for fixed windows of 50 bases, according to Blastz alignments. The values shown are L-scores where higher values mean better alignments.
First a Blastz pairwise alignment was created with human as the target and mouse as the query sequence. Then we scored each overlapping 50 base window by sliding 5 bases at a time. The scores computed here (s-scores) are given by... These window scores were filtered to only include those with more than 25 aligned bases. A set of non-overlapping windows with all of their bases in ancient repeats are held out as evolutionarily neutral and their scores make the neutral density used to compute the cumulative distribution function CDF so that CDF(s) = P(x < s) for a uniform random variable x. Thus the L-score for s is -log10( 1 - CDF(s)) and the L-scores are displayed on the track itself for all the windows.
Thanks to Mark Diekhans for scoring the windows and selecting out the ancient repeats. Krishna Roskin then created s-scores for these windows including the local background calculation. Ryan Weber computed the CDF for these s-scores, divided up the data by Blastz best mouse alignment, and created the 'sample' track with the detailed alignment display.