Abstract Introduction The cause of sporadic Alzheimer's disease (AD) remains unclear. Given the growing evidence that protein aggregates can spread in a “prion-like” fashion, we reasoned that a small population of brain cells producing such “prion-like” particles due to a postzygotic acquired mutation would be sufficient to trigger the disease. Deep DNA sequencing technology should in principle allow the detection of such mosaics. Methods To detect the somatic mutations of genes causing AD present in a small number of cells, we developed a targeted deep sequencing approach to scrutinize the genomic loci of APP , PSEN1 , and PSEN2 genes in DNA extracted from the entorhinal cortex, one of the brain regions showing the earliest signs of AD pathology. We also included the analysis of the MAPT gene because mutations may promote tangle formation. We validated candidate mutations with an independent targeted ultradeep amplicon sequencing technique. Results We demonstrate that our approach can detect single-nucleotide mosaic variants with a 1% allele frequency and copy number mosaic variants present in as few as 10% of cells. We screened 72 AD and 58 control brain samples and identified three mosaic variants with low allelic frequency (∼1%): two novel MAPT variants in sporadic AD patients and a known PSEN2 variant in a Braak II control subject. Moreover, we detected both novel and known pathogenic nonmosaic heterozygous variants in PSEN1 and PSEN2 in this cohort of sporadic AD patients. Discussion Our results show that mosaic mutations with low allelic frequencies in AD-relevant genes can be detected in brain-derived DNA, but larger samples need to be investigated before a more definitive conclusion with regard to the pathogenicity of such mosaics can be made.