Reference base composition versus C-to-T substitution frequencies in 1-nt 5′ overhangs in libraries from various Neanderthal specimens (N = 6). As some libraries show substantial levels of human DNA contamination, the analysis was restricted to putatively endogenous, deaminated fragments by requiring a C-to-T substitution at the 3′ end of sequences. Background noise was subtracted using the signal-to-noise ratio determined via comparison to a modern control sample (see Supplemental Table S2). For specimens for which more than one library was available, the library producing the largest number of DNA fragments with an inferred 1-nt 5′ overhang was chosen. Moderately significant positive or negative correlations were found for T (purple; one-tailed F-statistic linear regression on 1° and 4° of freedom: 17.81, P-value 0.01 [R2:0.8166]) and C (green; one-tailed F-statistic linear regression on 1° and 4° of freedom: 8.80, P-value 0.04 [R2: 0.6874]), but not for A or G (F: 0.44, P-value 0.55 [R2: 0.09875] and F: 0.89, P-value 0.40 [R2: 0.1825]).
