In popular culture artificial intelligence is more often associated with homicidal robots, self-aware computer systems hell bent on global destruction or defeated chess grandmasters than positive health outcomes.
But in the real world AI is a force for medical good according to Wolfgang Boomgaarden from PharmaInformatic, who told us his AI-based IMPACT-F system helps drugmakers identify candidates with the best oral bioavailability profiles during lead development.
“IMPACT-F is an in-silico method which solely needs the compound structures of interest to forecast oral bioavailability in humans” said Boomgaarden, explaining that the system is based on an extensive database of drug bioavailability information he put together.
“That knowledgebase took 10 years to develop. I applied different methods from AI technology such as neural networks or decision trees to develop predictive models. At the end I combined the best models and that yielded to the expert system IMPACT-F.”
Boomgaarden claimed the system is more accurate than animal model-based bioavailability predictions, explaining that: “Often animal drug-uptake is used to predict human outcome, but there are large differences in human and oral bioavailability.”
Cost saving
The other advantage for drug developers is cost according to Boomgaarden, who said: “Since it is a structure based approach, no additional costs such as assays, further experimental tests or compound synthesis are needed.”
At present Emden, Germany-based PharmaInformatic provides IMPACT-F as a service, using the technology to assess structural diagrams of candidate compounds provided by customers in the pharma, biotechnology and contract research sectors.
However, a licensing model is planned according to Boomgaarden, who said: “In future, customers can do the calculations in-house, so they do not have to disclose compound structures.”