The Challenge of Whole Exome Sequencing As a Molecular Diagnosis for ASD

Background:  The search for an objective diagnosis for autism spectrum disorder (ASD) is a major concern of the scientific community. The availability of new molecular tests increased the expectations for an etiologic diagnosis of ASD. Nowadays, it would be expected that these tests would allow a conclusive result in about 20-30% of the cases. Whole exome sequencing (WES) are being offered as an approach for ASD diagnosis, but because of the amount of data and difficulties in interpreting the clinical significance of the variants in most situations it is a challenge to prioritize variants and return a final conclusive report for the family. In this context to determine the actual power of the WES as a diagnostic test is imperative.

Objectives: To investigate the specificity and sensitivity of WES as a molecular test for ASD.

Methods: 40 individuals with ASD, 115 individuals with other clinical diagnosis than ASD and 609 higid individuals with more than 60 years old. The sequencing was made with Illumina’s platform and alignment were performed with bwa and variant calling for ASD and 115 control samples (OSC) together with Unified Genotyper from GATK, and the 609 control(HSC) were called separately with the same software. Annotation was performed with ANNOVAR and internal pipelines. The variants were filtered for 243 genes present in SFARI database from the three first categories and syndromic. Variants of Low Quality, outside exonic regions, genotype calling under 99, and minor alelle frequency inferior to 0,3 were removed from the analysis. Variants were separated by frequency in 3 groups, above 0.05, between 0.05 and 0.01 and below 0.01. Each individual had their variants accounted and prioritized, loss-of-function (LoF) variants (frameshift, stopgain/loss and splicing site) were divided into two groups, one considering all the variants in the SFARI genes, and other taking into account the ExAC LoF intolerance genes, creating a subset of 151 genes from SFARI. Missense variants were prioritized using SPRING software, and only those variants with p-value below 1.10^-12, were included. False positive and false negative rates were calculated for each type of variants and ROC curves were plotted.

Results: There were no difference in the mean number of LoF (ASD: mean 6.01, sd 6.31; OSC: mean 6.90, sd7.10; HSC: mean 5.92, sd5.06), ExAC LoF (ASD: mean 3.69, sd 4.01; OSC: mean 4.17, sd 5.01; HSC: mean 3.60,sd 3.93), and missense variants between the groups (ASD:mean 6.07, sd 3.34; OSC: mean 6.44, sd 3.66; HSC: mean 6.59, sd 3.49). The false positive rates and false negative rates for all the groups were above 50%. The area under de ROC curve were very close to 0,5 in all occasions.

Conclusions:  The prioritized methods chosen above were not sufficient to discriminate between individuals with ASD and controls. Further studies must be completed in order to create an effective analysis for WES in patients with ASD. Research supported by FAPESP-CEPID, CNPq, INCT.