Distribution in ratio of actual-to-predicted genomic extent, as function of gene size. Genscan is given a sequence containing the cDNA-aligned exons and all intervening introns. When multiple genes are predicted, the one with the longest genomic extent is taken, as long as there is at least some overlap with the actual exons. Genscan performs well for small genes, but its performance is severely degraded for genes above 100 kb.