Abstract: Background: Homozygous deletions are rare in cancers and often target tumor suppressors. Reliable identification of homozygous deletions in primary tumors has historically been hindered by the admixture of non- aberrant cells. Materials and Methods: Using recently developed bioinformatics approaches, we are performing a systematic screen of homozygous deletions in a compendium of over 10,000 human cancers, aiming to identify known and novel tumor suppressors. Recurrent homozygous deletions occur either over tumor suppressors or over fragile sites, regions of increased genomic instability. We construct a statistical model that separates fragile sites from regions showing signatures of positive selection for homozygous deletions that therefore are likely to contain tumor suppressors. From the patterns of the homozygous deletions, the patterns of mutations in COSMIC and the literature, we subsequently identify candidate tumor suppressors. Results: Results on a pilot series of 2218 primary tumors across 12 human cancers identified 96 genomic regions recurrently targeted by homozygous deletions. This pilot screen identified 12 established tumor suppressors and discovered 24 candidate tumor suppressors. We observe homozygous deletions targeting CPEB3, mostly in lung cancer. CPEB3 encodes a sequence-specific RNA-binding protein that has been shown to transcriptionally repress EGFR. Therefore, our results suggest that, in addition to amplification of EGFR, lung cancers can also adopt homozygous deletion of CPEB3 as an alternative mechanism to overexpress EGFR. We also found homozygous deletions affecting CASP3 and CASP9, or regions close by, in various cancer types. CASP8, another member of the caspase family, is a known tumor suppressor in breast cancer. Both CASP8 and CASP9 induce apoptosis through cleavage of CASP3. It is therefore likely that complete inactivation of any of these three caspases may abrogate apoptosis. For several other genes with recurrent homozygous deletions, prior evidence, often recent, was found in the literature that these are indeed tumor suppressors (including FAT1, BIRC2/BIRC3, MGMT and USP44). Many other genes are novel, with no prior evidence of a tumor suppressive role (including KAT6B, BAZ1A, USP34, LINC00662 and KIAA1551). Several of these genes play a role in e.g. DNA repair or apoptosis, processes where one would expect to find tumor suppressors. Conclusions: Through copy-number meta-analysis of large-genomic datasets, our study identifies tumor suppressors through recurrent homozygous deletions, significantly advancing the landscape of driver genes in human cancer.