Methods

The human/mouse alignments produced by blastz was analyzed by counting the number occurrences of aligned human and mouse base pairs in categorized sites. The number of substitutions per site from the ancestral DNA was estimated using the REV model. Categorization was done based on the type of site, GC content, and if CpGs were included. Results created for all possible combinations of these attributes.

The following types of sites considered:

• ancestral repeat sites - Bases belonging to aligned transposons identified by RepeatMasker.
• context-conserved 4D sites - The third base of a 4D codon where the first two bases are conserved in the alignment.
• context-unconserved 4D sites - The third base of a 4D codon in human or mouse where the first two bases are not conserved in the alignment.
• all 4D sites - A both context-conserved and context-unconserved 4D sites.
• all aligned sites - All bases in the alignment.

GC content was computed from the aligned bases in 100kb, non-overlapping windows of the human genome. This was done using either the human or mouse DNA. The windows were then partitioned into three ranges that give a uniform distribution of the human GC content windows. This resulted in the following ranges:

• Low: GC <= 36.2%
• Medium: 36.2% < GC <= 41.2%
• High: 41.2% < GC

Sites were identified as being CpG if they contained either base on a CG pair on the positive strand. Note that this also includes CpGs on the negative strand in opposite direction.