Table 1.

Overview of outcomes for the top nine competitors

Rank	Method	Language	Spearman correlation (%)	AUC extreme modulations (%)	Recovered gene knockdowns (%)	MSE	Mean inter-replicate correlation (%)	Time per plate (s)
1	Random forest regressor	Java	66.5	91.5	77.2	1	45.2	14.5
2	Gaussian mixture model	C++	65.4	91.4	75.7	1.1	43.1	4
3	Modified k-means	C++	64.6	91.2	77.8	2.2	41.9	10.5
4	ConvNet	Python/C++	64.8	91	76.6	2.4	41.8	25
5	Gaussian mixture model	Python/C++	64.6	90.9	75.7	1.3	41.9	36
6	Modified k-means	Python/C++	64.3	90.2	70.8	1.1	40.6	11.5
7	Boosted tree regressor	Python	64.5	91.1	77.2	1.7	41.9	50.5
8	Modified k-means	Python	65.1	90	69	1.2	43.7	35.5
9	Other	Java	63.9	89.9	75.1	1.5	39.6	4.5
BM	k-means	Matlab	63.2	89.2	73.9	3	38.9	247

Note: All the values are based on the holdout dataset; see the main text for the meaning. The maximum and minimum values in each column are in bold.