Somatic mosaicism exists among tissues widely and would mark circulating cell-free DNA (cfDNA) as DNA fragments released by lytic cells from distinct tissues into the blood. By investigating the alignment pattern of sequencing reads from whole genome sequencing on genomic DNA of different tissues, we found the reads distribution forms tissue-specific patterns on some regions, as a result of somatic mosaicism. We then utilized this indication to construct a tissue-of-origin mapping model and evaluated the predictive performance on WGS data from tissue and cfDNA. In total, 1,545 tissue samples involving 13 cancer types were included, and the performance of identification of tissue-of-origin achieved specificity of 82% and sensibility of 80%. Furthermore, a total of 30 cfDNA samples involving lung cancer, liver cancer, and healthy control were analyzed to indicate their nidus’ tissue-of-origin with specificity and sensibility both at 87%. Our results show that reads distribution of whole genome sequencing could be used to identify the tissue-of-origin of cfDNA samples with high accuracy, suggesting the potential application of our model on early tumor detection and diagnosis.