Table 2 Top features according to information gain and information gain ratio (excluding combination features)
From: Improved identification of conserved cassette exons using Bayesian networks
Rank | Feature | Information Gain | Feature | Information Gain Ratio |
|---|---|---|---|---|
1 | Length of best alignment in the upstream intron flank | 0.169 | Abundance of GA in exon | 0.172 |
2 | Upstream intron flank conservation | 0.169 | Density of single stranded ESEs in exon | 0.151 |
3 | Identity of best alignment in the upstream intron flank | 0.142 | Exon identity | 0.128 |
4 | Downstream intron flank conservation | 0.138 | Average of positive NI scores in exon | 0.118 |
5 | Length of best alignment in the downstream intron flank | 0.138 | Length of best alignment in the upstream intron flank | 0.117 |
6 | Exon identity | 0.120 | Density of AC in exon | 0.115 |
7 | Identity of best alignment in the downstream intron flank | 0.088 | Average of negative NI scores in exon | 0.112 |
8 | Exon length | 0.080 | Density of CT in exon | 0.111 |
9 | Matches in 12-mer near 3'ss | 0.066 | ESE density in exon | 0.104 |
10 | Symmetry | 0.042 | Length of best alignment in the upstream intron flank | 0.103 |