Comparison of speed and accuracy of MLgsc versus FrameBot for the nitrogenase gene <i>nifH</i>.

(a) For MLgsc, a query was considered not classifiable if any node in the corresponding output line had an evidence ratio (ER) below the cutoff. The use of evidence ratio do detect sequences that are not confidently classified is described in section Procedure, subsection Validation. (b) A FrameBot classification was evaluated by examining the lines starting by STATS in the *framebot.txt output file:, which contains the predicted genus name and the test sequence's ID, for example in STATS 454423|B|1|1594_2487_L23514 Nostoc_commune_UTEX_584_nitrogen_fixation_protein_nifUAAA21838 the predicted genus is Nostoc and the query ID is AAA21838. Using the ID to look up the genus in the reference file, we consider the prediction correct if the two genera match. (c) FrameBot command: java-jar dist/FrameBot.jar framebot-N-o test_nifH refset/nifh_prot_ref.fasta ENA_nifH_full_cleanHdr.dna (d) MLgsc command: mlgsc ENA_nifH_full_cleanHdr_orf800_bestORF.pep NifH_ref_clean_train.bcls Comparison of speed and accuracy of MLgsc versus FrameBot for the nitrogenase gene nifH.