Table 5. For each search term used to identify putatively important conserved protein domains, we show the number of domain descriptions that contain this term for various categories including: (i) the starting set, (ii) the input to the Random Forest model, (iii) the top 50 and (iv) the top 20 features after model fitting.

Note that column entries will not sum up to the n depicted at the top of each column as many descriptions contain multiple search terms.

	Starting	Model
Search term	(n = 371)	(n = 206)	Top 50	Top 20
integrase	101	72	24	13
excisionase	5	4	2	0
recombinase	72	52	28	17
transposase	143	70	14	3
lysogen	23	10	2	1
temperate	11	10	3	0
parA |ParA |parB |ParB	65	29	7	0