Cross-level analysis of molecular and neurobehavioral function in a prospective series of patients with germline heterozygous PTEN mutations with and without autism

Table 3 Hierarchical regression and support vector machine models with PTEN mutation status and protein assays predicting neurobehavioral measures

	Hierarchical regression				Support vector machine
	Predictors retained beyond PTEN mutation status	PTEN mutation	Protein levels		PTEN mutation and protein levels
		R²	ΔR²	ΔF (p)	Fivefold cross-validation R/R²
Full-scale IQ	PTEN, total AKT	< .01	.17	5.90 (.005)	.33/.11*
Verbal IQ	PTEN, total AKT, MnSOD	< .01	.16	4.97 (.004)	.35/.12*
Non-verbal IQ	EIF2A	< .01	.08	4.73 (.034)	.28/.08*
Language composite	EIF2A	< .01	.09	5.57 (.022)	.27/.08*
Frontal-subcortical composite	PTEN, p27	< .01	.16	5.56 (.006)	.26/.07*
Attention composite	p27	.02	.07	4.72 (.034)	.14/.02
Working memory	–	.02	–	–	.37/.14*
Processing speed	–	.01	–	–	.20/.04
Motor skills composite	p27	.01	.10	6.48 (.014)	.01/.01
Internalizing problems	–	< .01	–	–	.24/.06
Externalizing problems	P-AKT, P-S6, S6	.02	.20	4.67 (.006)	.47/.22**
Autism traits	–	.03	–	–	.17/.03
Repetitive behavior	Total P-ERK, P-S6	< .01	.20	4.66 (.006)	.22/.05
Adaptive behavior	–	.01	–	–	.10/.01

Hierarchical regression models included PTEN mutation status in step 1 and all protein levels in step 2, with specific predictor entry set at p < .05 and removal at p > .10. Support vector models included PTEN mutation status and all protein assays in separately predicting each neurobehavioral variable. Support vector model fivefold cross-validation R significance based on full sample. *p < .05, **p < .001

ISSN: 2040-2392