hypothesized that small CNVs (between 1 and 50 Kb), which are below the resolution of most commercial microarrays and and whose detection still has limitations for its detection detection through NGS, could contribute to the phenotype in a proportion of cases. As a first step to address this hypothesis, it was performed the methodology of custom aCGH 60K, covering a total of 269 ASD candidate genes, in order to select potentially pathogenic CNVs among 98Brazilian patients with idiopathic ASD. With this initial screening, the prevalence of potentially pathogenic CNVs obtained was 9%, with 20% of them characterized as small. The subsequent analysis was performed using the 180K custom aCGH methodology, which covered a total of 1527 TEA candidate genes. A total of 63 patients with autism were analyzed with this new array. From these hybridizations, the prevalence of potentially pathogenic CNVs obtained was 12.7%, with 62.5% of them classified as small. This detection rate is quite significant, particularly considering that the sample of patients used was prescreened, in order to exclude patients with the most prevalent CNVs in ASD in the regions 15q11-q13, 16p11.2 and 22q13.3. The last approach used in this work was to compare the detection of CNVs by the methodology of aCGH, gold standard reference for CNVs detection, with the next generation sequencing (NGS).Data from 9 patients obtained by NGS were analyzed using NextGene and XHMM software. The softwares, however, presented discrepant results among themselves and little overlap with the data of aCGH 180K, of 38.9% and 50%, considering NextGene and XHMM respectively. These results suggest that the customized aCGH represents a promising approach for the detection of small CNVs and that these, in turn, can contribute to the risk of ASD in at least 6,3 % of cases.