Once the data is generated and crunched, then what? Can we accurately rank our compounds? Do we have confidence in that ranking?
SimuGen’s raison d’etre is to include a third stage after biological modeling and data modeling—which Wills calls “decision modeling.” Its web-based algorithms are designed to be agnostic: To seamlessly combine into a single model gene and protein-expression data, cell viability and impedance assays, and high-resolution microscopic imagery, for example, “in an easy way that everybody can understand, and in a way that helps us make a decision,” he explained.
Researchers can implement their own model, utilizing their favorite cell types, assays, and endpoints. The software allows researchers to combine those into a screen that can be sold or used in-house.
Others may want to utilize SimuGen’s own models, currently focused on six major liver toxicity endpoints. “Liver is by a long stretch the biggest reason for drug failure on the market,” noted Wills. Experiments can be performed in-house, or by the company’s service partners, and then analyzed for six different endpoints.
The software will perform quality control and will rank the chemicals, showing which toxicities are occurring and at what concentration toxicity begins to occur, he explained. “It will then also show you at a global level how all the chemicals are looking relative to each other, so that you can start spotting patterns of chemicals that are behaving similarly in terms of their toxic profiles.”
The latest version of SimuGen’s software, Wills said, includes the ability to weight models—for example, downgrading certain endpoints relative to others. “Phospholipidosis should be far less of a flag compared with carcinogenesis.” This version also allows thresholds to be included. If blood levels of a compound aren’t likely to be seen at greater than 10 micromolars, “as a rule any toxicity we pick up at 100 micromolars is really irrelevant.”