Metagenome Fragment Classification Using 𝑁-Mer Frequency Profiles

Table 3

Comparison of the top 10 reads from the naive Bayes analysis of the Sargasso Sea set for 9 mers and 15 mers and a side-by-side comparison with MEGAN results. There are 7 common strains between the naive Bayes sets substantiating their presence in the sample. Not all NBC “best matches” are found in MEGAN (indicated by “None”), and this can be due to “no hits” or to not having that strain in the database. An interesting NBC find is that Trichodesmium erythraeum has been found to compose 0.6% of the sample. It has been extensively found in the Sargasso Sea, but no prior methods show this presence in the Sargasso Sea data set.

9 mers			15 mers
High-strain content in sample (genome size of both sides)	No. of reads	No. of MEGAN reads	High-strain content in sample	No. of reads	No. of MEGAN reads
Burkholderia 383 (9.3 M)	693	514	Burkholderia 383 (9.3 M)	2044	514
Burkholderia Cenocepacia AU 1054 (14.6 M)	684	13	Clostridium Beijerinckii NCIMB 8052 (12 M)	1698	2
Clostridium beijerinckii NCIMB 8052 (12 M)	623	2	Shewanella ANA-3 (10.3 M)	989	186
Shewanella ANA-3 (10.3 M)	562	186	Trichodesmium erythraeum IMS101 (15.6 M)	584	2
Trichodesmium erythraeum IMS101 (15.6 M)	533	2	Flavobacterium johnsoniae UW101 (12.2 M)	481	10
Burholderia xenovorans LB400 (19.6 M)	404	None	Sorangium cellulosum So Ce 56 (26 M)	309	None
Shewanella MR-4 (9.4 M)	329	14	Shewanella oneidensis MR-1 (10.4 M)	297	78
Burholderia ambifaria/cepacia AMMD (15 M)	265	91	Shewanella MR-4 (9.4 M)	245	14
Alkaliphilius metalliredigens QYMF (9.8 M)	261	None	Burkholderia cenocepacia HI2424 (15.5 M)	219	102
Shewanella MR-7 (9.6 M)	250	26	Shewanella MR-7 (9.6 M)	206	26
Acidobacteria bacterium Ellin345 (11.6 M)	187	None	Burkholderia xenovorans LB400 (19.6 M)	198	None