Table 2.

Comparison of estimated and measured runtime

Predictor	Coefficients of the linear fit into the measured runtime values	Runtime estimated using linear fit for the human proteome # sequences = 20,193 average length = 561	Runtime estimated using linear fit for the TESTsmall dataset # sequences = 432 average length = 332	Runtime measured for the TESTsmall dataset # sequences = 432 average length = 332
fDETECT	a = 0.12180127	0.71 h	0.90 min	0.88 min
	b = 8.2653E-06
PPCpred	a = 88.742223	76.15 days	27.49 h	28.35 h
	b = 0.42263874
Crysalis	a = 0.07497269	0.73 h	0.77 min	0.77 min
	b = 9.814E-05
PredPPCrys	a = 204.206124	79.12 days	34.04 h	34.48 h
	b = 0.23944459

The analysis covers the four methods that predict the four steps of the protein production and crystallization process. The second column shows coefficients of a linear fit into the measured values of the runtime and protein sequence length on the TESTsmall dataset, i.e., runtime = a*sequence_length + b. The total runtimes estimated with that linear fit for the proteins in the complete human proteome and from the benchmark dataset are listed in columns three and four, respectively. The right-most column shows the total runtime that was empirically measured on the TESTsmall dataset