Table 2.

Computational models of yECitrine fluorescence based on 5′-UTR structure

Name	Region 1	Region 2	Model	Correlation coefficient	Predicted residual sum of squares	Residual sum of squares
CYC1Model1	[−166, −45]	[−37, 6]	ln( f ) = 0.0986 × ΔG1 + 0.1253 × ΔG2 + 0.5004	0.7809	0.3016	0.01701
TEFModel1	[−137, −7]	[−6, −1]	ln( f ) = 0.1042 × ΔG1 + 41.5185 × ΔG2 − 0.6856	0.5922	1.5128	1.5226
GPDModel1	[−115, −98]	[−53, 19]	ln( f ) = 2.3378 × ΔG1 + 0.1227 × ΔG2 − 1.4524	0.8340	1.2294	0.02174
CYC1Model2	[−105, −95]	[−53, −5]	ln( f ) = 1.1331 × ΔG1 + 0.0936 × ΔG2 − 0.1545	0.8600	0.1904
TEFModel2	[−93, −87]	[−32, −8]	ln( f ) = 106.9974 × ΔG1 + 0.3197 × ΔG2 + 0.4363	0.9100	0.2278
GPDModel2	[−126, −99]	[−76, −4]	ln( f ) = 0.6411 × ΔG1 + 0.1221 × ΔG2 + 1.2860	0.9536	0.2264

Indicated regions are measured relative to the first nucleotide of the start codon. The correlation coefficient was computed for all data available at the time of model training. The PRESS was computed with the hat matrix after regression. The residual sum of squares was computed for CYCModel1, TEFModel1 and GPDModel1 with the natural log of the data from pCYC1₂xYFP, pTEF₂xYFP and pGPD₂xYFP, respectively.