Microsoft Research has released source code for software tools to advance AIDS vaccine research, with the company’s spam filter playing a role in the effort. Available for free download on Microsoft’s CodePlex site, Microsoft is hoping to entice other scientists to use and build upon the tools to expedite development of a vaccine. The tools are intended to help researchers harness computing power to more quickly identify crucial elements of an effective cellular vaccine, Microsoft said. Microsoft researchers began work in AIDS research in February 2005 after discovering that machine-learning technology, including methods for spam-filtering, had applications for vaccine research. The vaccine work combines technologies including graphical models to comb through HIV strains to find genetic patterns necessary for training patient’s immune system to fight the virus, according to Microsoft. Code for four tools is being released. Researchers can download pre-compiled programs and run them on their own computers, providing more control, Microsoft said. Or, they can download source code and compile the applications themselves, enabling scientists to modify and build on the code.The tools let researchers sift through complex immune and viral genetic information from a population of infected people to pinpoint key strings of amino acids that must be present in a vaccine for it to be effective. Microsoft also is posting the actual tools on the Microsoft.com Web site. Specific tools include: * PhyloD, for incorporating a phylogenetic or evolutionary tree into analysis of patterns. This tool looks for correlations between a person’s human leukocyte antigen (HLA) system and the virus that infected that person. * Epitope Prediction, using machine-learning to relate Microsoft’s spam-filtering technology to scan proteins for likely epitopes in people with any HLA type.* HLA Assignment, for finding epitopes more accurately. * HLA Completion, to help scientists address the hierarchy of any immune system’s HLA types. Software Development