Improving stability and understandability of genotype-phenotype mapping in Saccharomyces using regularized variable selection in L-PLS regression

Table 1 An overview of model parameters and complexity

Phenotype	f	d	α	No. of	RMSE	No. of selected	No. of selected
				components		genes	background variables
Melibiose 2% Rate	0.1	0.99	0.73	5	0.61	30	15
Melibiose 2% Efficiency	0.1	0.99	0.65	4	0.60	30	14
Cupper chloride 0.375mM Rate	0.5	0.99	0.78	6	0.57	31	15
Cupper chloride 0.375mM Efficiency	0.1	0.90	0.51	9	0.52	30	14
NaCl 0.85M Rate	0.1	0.95	0.62	1	0.61	31	14
NaCl 1.25M Rate	0.5	0.95	0.85	2	0.80	30	15
NaCl 0.85M Efficiency	0.1	0.95	0.77	4	0.71	30	15
NaCl 1.25M Efficiency	0.1	0.99	0.67	5	0.60	30	15
Maltose 2% Rate	0.1	0.99	0.63	5	0.60	31	15
Maltose 2% Efficiency	0.1	0.90	0.54	7	0.50	30	15
Galactose 2% Rate	0.1	0.95	0.75	7	0.50	30	16
Galactose 2% Efficiency	0.1	0.95	0.56	7	0.61	30	15
Heat 37^oC Rate	0.1	0.99	0.65	6	0.58	30	15
Heat 40^oC Rate	0.1	0.99	0.78	6	0.82	30	15
Heat 37^oC Efficiency	0.1	0.99	0.51	1	0.59	31	14
Heat 40^oC Efficiency	0.5	0.90	0.62	8	0.67	30	15
Sodium arsenite oxide 3.5mM Rate	0.1	0.90	0.59	8	0.54	30	15
Sodium arsenite oxide 5mM Rate	0.1	0.99	0.66	5	0.62	30	15
Sodium arsenite oxide 3.5mM Efficiency	0.1	0.99	0.73	3	0.55	31	14
Sodium arsenite oxide 5mM Efficiency	0.1	0.90	0.51	2	0.63	31	14

The suggested approach select the model parameters at each level of 10-fold cross validation, hence the measure of central tendency is used to summaries the results. For each fitted model mode of selected step length (f), mode of rejection level (d), mode of model complexity (α and no. of components), mean of RMSE on test data, number of selected genes and number of selected background variables are listed.

ISSN: 1471-2105