An important aspect of cancer therapy is the specificity of the treatment. Significant progress has been made in drug design. However, gene therapy by direct targeting of oncogenes has not yet met equal success. Among the most valuable tools are meganucleases, also known as homing endonucleases (HE). These enzymes recognize and cleave long (around 20bp) DNA sequences and have been shown to induce gene repair in vivo efficiently. The mechanism involves induction of homologous recombination (HR) by creating a double strand break (DSB) of DNA. HEs also have the significant advantage of being amenable to redesign for new target sequences. Applications have been demonstrated, for example, in the treatment of xeroderma pigmentosum, a condition associated with skin carcinomas. Unfortunately, inducing HR by DSB tends to produce genetic instabilities. A safer alternative would be to use single-strand-cleaving enzymes (nickases). However, this must be done retaining sequence specificity, a property unavailable in natural enzymes. Consequently, engineering is necessary. A promising candidate for redesign is the monomeric HE I-DmoI, but details of the mechanism, namely the role of metals in the active site and the precise timing of strand cleavage, must still be unraveled. X-ray and mutagenesis experiments performed at CNIO provided significant, although incomplete, evidence on those crucial aspects. Experimentally-supported theory and computational modeling can help clarify remaining incognitas, such as the strand preference of the enzyme. We will use state-of-the-art computational methods (ab-initio molecular dynamics, hybrid Quantum Mechanics/Molecular Mechanics and molecular modeling), to gain insight into aspects of the mechanism not seen in experiments, including the timing of strand cleavage. This knowledge could eventually lead to the design of a more effective and specific HEs to be used in gene therapy of skin cancer, and hopefully of other types.