Figure 3

D _N / d _S cumulative frequency distribution of references, gene predictions and pseudogene predictions sets. Benchmark, predicted genes, pseudogenes (with one mutation) and pseudogenes (with accumulated mutations) sets exhibit a median d_N/d_S of 0.15, 0.18, 0.22, 0.47, respectively, compared to their human functional orthologues. While the d_N/d_S distribution of pseudogenes with accumulated mutations sets is clearly shifted upwards to the theoretical value of 0.57 (average between 1.0 for no selection and 0.15 for selection from the benchmark set), the pseudogene set with one mutation is not significantly shifted suggesting this set may contains spurious pseudogene prediction. Predicted and benchmark gene sets have a similar d_N/d_S cumulative frequency distribution indicating comparable selective constraints level.