OBJECTIVES
Autism is a neurodevelopment disorder with unclear etiology. High heterogeneity is one of the main issues in the etiological studies. This study explores the relationship between RELN signaling pathway related genes (RELN, VLDLR, LRP8, DAB1, CDK5, FYN) and language development of autism patients based on a cluster analysis model which is established to reduce the heterogeneity.
METHODS
Autism children were recruited from 5 different medical/autism training institutes from Hunan, Shandong, and Henan provinces, and were divided into 2 parts according to the recruitment time: The first part was the training sample, which was recruited from October 2006 to May 2011, and the second part was the validation sample, which was recruited from July 2011 to May 2012. A two-step cluster analysis was performed to cluster 374 Chinese Han autism patients into different subgroups based on 2 parameters: Onset age of the first word and interval from the first word to the first phase. A Bayes discriminatory equation was established followed the cluster results. Then we used this equation to divide another 310 autism children into prior defined subgroups. After the genotyping data was screened, a single marker case-control association study was conducted.
RESULTS
The cluster analysis clustered 374 samples into 3 subgroups. Onset ages of the first word in the Group A were (11.83±4.37) months and intervals from the first word to the first phase were (24.55±8.67) months; onset ages of the first word in the Group B were (12.17±3.46) months, intervals from the first word to the first phase were (7.07±3.79) months; onset ages of the first word of Group C were (30.94±7.60) months, intervals from the first word to the first phase were (4.73±4.80) months. The established equations based on the cluster analysis were YA=-14.442+0.525X1+0.810X2, YB=-4.964+0.477X1+0.264X2, YC=-19.843+1.175X1+0.241X2. Cross validated analysis showed that the false rate of the equation was 3.8%. A total of 341 single nucleotide polymorphism (SNP) in 6 genes passed the quality control. Before divided subgroups, none of these SNPs reached the significant P value (P>2.44×10-5, Bonferroni adjustment). However the result showed that rs1288502 of LRP8 in Group B was significantly different from the control group (P=6.45×10-6).
CONCLUSIONS
Based on the cluster analysis of language development, we could establish a discriminatory equation to reduce heterogeneity of autism sample. The association test indicates that LRP8 genein RELN signaling pathway is related to a particular type of language development of autism patients.