Table 1 Performance metrics of the models on the test sets.

	σ⁷⁰ specific promoters		σ^B specific promoters		σ^F specific promoters		σ^W specific promoters
	WT genotype		σ^B genotype		σ^F genotype		σ^W genotype
	Mean	Std.	Mean	Std.	Mean	Std.	Mean	Std.
Spearman’s rho	0.574	0.003	0.565	0.002	0.497	0.002	0.234	0.050
Weighted ACC	0.230	0.005	0.230	0.005	0.210	0.004	0.136	0.015
Weighted MAE	1.609	0.007	1.652	0.017	1.919	0.052	2.504	0.100
MAE
y = 0	2.866	0.150	2.397	0.068	2.891	0.107	4.797	0.240
y = 1	1.845	0.105	1.630	0.070	2.097	0.117	4.022	0.141
y = 2	1.372	0.054	1.071	0.028	1.507	0.101	2.739	0.287
y = 3	1.284	0.058	1.197	0.035	1.294	0.056	1.613	0.329
y = 4	1.241	0.045	1.368	0.058	1.324	0.095	1.090	0.122
y = 5	1.324	0.023	1.407	0.055	1.357	0.049	0.587	0.480
y = 6	1.401	0.038	1.333	0.036	1.258	0.068	0.951	0.111
y = 7	1.362	0.045	1.315	0.051	1.572	0.119	1.790	0.189
y = 8	1.316	0.055	1.507	0.065	2.102	0.146	2.711	0.246
y = 9	1.533	0.032	1.730	0.116	4.010	0.135	3.774	0.205
y = 10	2.158	0.047	3.224	0.165	1.704	0.320	4.646	0.298

			σ^F genotype		σ^B genotype		σ^B genotype
ROC AUC			0.694	0.004	0.652	0.004	0.615	0.010

			σ^W genotype		σ^W genotype		σ^F genotype
ROC AUC			0.691	0.002	0.643	0.004	0.635	0.006

			WT genotype		WT genotype		WT genotype
ROC AUC			0.665	0.004	0.632	0.003	0.635	0.006

Weighted metrics are used for accuracy and mean absolute error to account for class imbalance. For each performance, the mean and standard deviation (Std.) are given obtained by training multiple models in a five-fold set-up for the test set. Mean absolute errors for each of the sample classes (y = 0–10) are given. ROC/PR AUC is used for the binary classification problem. ROC AUC represent a perfect model at AUC = 1. (ACC accuracy, MAE mean absolute error, AUC area under the curve, ROC receiver operating characteristic, WT wild-type, σ sigma factor). Source data are provided as a Source Data file.

Quick links

Search