Ovarian cancer is the fourth most common cancer among women and existing treatment is not routinely curative. One new strategy for cancer therapy is the selective delivery of TNFalpha to tumors via adenovirus vectors. We have tested the combination of two modifications to adenovirus vectors designed to limit delivery to tumors, capsid modification and expression control. To target alpha(v)beta(3/5) integrin receptors that are highly expressed in tumor and sparsely expressed in the epithelial layer of peritoneum, we modified the capsid fiber and penton base to remove native receptor binding and incorporated an RGD-4C motif in the fiber knob (Ad.PB*F*RGD). This vector exhibits effective gene transfer in all of the alpha(v)beta(3/5)-positive ovarian cancer cells tested in vitro and in vivo. Importantly, the Ad.PB*F*RGD vector is able to transduce ovarian tumor nodules and avoid infecting the normal mesothelial cells that line the intraperitoneal space following intraperitoneal administration. To further increase selectivity, different promoters were incorporated into the capsid-modified vector to confer the expression of the hTNFalpha therapeutic gene. We analyzed both constitutive (CMV or RSV) and potentially tumor selective promoters (MUC-1, E2F or hTERT) in terms of efficacy, selectivity and safety. TNF-expressing Ad.PB*F*RGD vectors containing the MUC-1 promoter showed anti-tumor activity in two ovarian cancer xenograft models (Caov3 and Igr-ov1) with little evidence of toxicity or systemic TNF. The data indicate that combination of capsid modification and transcriptional regulation of expression is a promising strategy for development of a new ovarian cancer treatment.