A hybrid model based on general regression neural network and fruit fly optimization algorithm for forecasting and optimizing paclitaxel biosynthesis in Corylus avellana cell culture

Table 2 Importance (according to the sensitivity analysis) and optimal levels of the different factors including cell extract (CE), culture filtrate (CF) concentration levels [% (v/v)], fungal elicitor adding day and harvesting time (day) for achieving maximum growth, paclitaxel biosynthesis and its secretion in Corylus avellana cell culture by the optimization analysis using FOA and genetic algorithm (GA) on developed general regression neural network-fruit fly optimization (GRNN-FOA) models

Criteria	Variable	Importance value (according to VSR^a)	Optimal level		Output optimal
Criteria	Variable	Importance value (according to VSR^a)	FOA	GA	FOA	GA
Dry weight (g l⁻¹)	CE concentration level	0.0068	4.26	5.34	12.57	12.18
	CF concentration level	0.0053	0.54	0.71
	Adding day	0.0237	16.33	15.62
	Harvest time	1.0022	20.58	20.86
Intracellular paclitaxel (µg g⁻¹DW)	CE concentration level	0.8904	4.12	3.45	19.26	18.53
	CF concentration level	0.6751	5.84	5.68
	Adding day	0.4257	15.72	16.97
	Harvest time	0.2443	20.34	20.41
Intracellular paclitaxel (µg l⁻¹)	CE concentration level	0.7458	4.43	5.07	224.78	213.78
	CF concentration level	0.4406	5.69	5.46
	Adding day	0.4083	16.09	16.79
	Harvest time	0.3961	20.47	21.09
Extracellular paclitaxel (µg l⁻¹⁾	CE concentration level	0.2862	4.58	4.73	152.15	141.11
	CF concentration level	0.7519	4.71	5.06
	Adding day	0.1893	15.91	16.19
	Harvest time	0.9477	22.06	22.86
Total yield of paclitaxel (µg l⁻¹)	CE concentration level	0.6891	4.29	4.97	372.89	369.04
	CF concentration level	0.6043	5.38	5.01
	Adding day	0.0943	17.00	16.53
	Harvest time	0.2018	20.68	20.16
Extracellular paclitaxel portion (%)	CE concentration level	0.1003	4.62	5.06	50.36	49.63
	CF concentration level	0.0622	4.91	4.97
	Adding day	0.1409	16.59	17.03
	Harvest time	0.4224	22.66	21.98

^aRelative indication of the ratio between the variable sensitivity error and the error of the model when all variables are available

ISSN: 1746-4811