Mutagenesis

We used the Mutagenesis dataset of 230 molecules, where each example is repeated 10 times to make more accurate timings possible and to have a better idea of the effect on larger datasets. We used three different language biases. 'small' is a language bias that was chosen so as to generate a limited number of refinements (i.e., a relatively small branching factor in the search lattice); this allows us to generate query packs that are relatively deep but narrow. 'medium' and 'large' use broader but more shallow packs. Table 3 summarises the number of queries and the number of frequent queries found for each level in the different languages.

Table 3: Number of queries for the Mutagenesis experiment with WARMR.

Mutagenesis
Level	small		medium		large
	Queries	Frequent	Queries	Frequent	Queries	Frequent
1	8	5	37	26	45	31
2	60	14	481	48	1071	211
3	86	24	688	114	3874	1586
4	132	31	699	253
5	37	21	697	533
6	29	18	1534	1149
7	23	15	-	-
8	17	12	-	-
9	4	4	-	-