Skip to main content

Table 4 Translation initiation site prediction performance of the new gene prediction algorithm (Neural Net) and MetaGene according to »reliable annotation subsets« (A subset of »verified genes« from »EcoGene« for Escherichia coli [28], all non-y genes of the Bacillus subtilis GenBank annotation and the »PseudoCAP« annotation of Pseudomonas aeruginosa [29]). TIS prediction sensitivity and correctness were measured on artificial 700 bp fragments that were randomly excised from each test genome to 5-fold coverage. Mean and standard deviation over 10 replicates per species are shown.

From: Gene prediction in metagenomic fragments: A large scale machine learning approach

  SENSITIVITY TIS TIS CORRECTNESS
Species Neural Net MetaGene Neural Net MetaGene
Bacillus subtilis 73.4 ± 1.79 62.1 ± 1.43 84.1 ± 0.51 70.2 ± 0.64
Escherichia coli 80.0 ± 0.68 75.1 ± 0.61 86.6 ± 0.57 77.5 ± 0.67
Pseudomonas aeruginosa 68.0 ± 0.22 79.7 ± 0.44 80.7 ± 0.20 83.7 ± 0.36