Regulators want to compile data to reduce animal testing for predictive toxicology, but IP could be a roadblock. Second in a series.

In the US, the new Standard for Exchange of Nonclinical Data (SEND) model could make it possible to replace animals with software for predictive toxicology tests. The purpose of SEND is to standardize the way researchers save nonclinical study data, so in theory that data could be read and organized by anyone with access.

Following the logic, it would be possible to compile all the data from every toxicology study into one massive data repository. Data standardization makes it easy to include every pharma or non-pharma chemical toxicity study into one big library, which would give the computer thousands of chemical structures matched with their toxic effects.

“The purpose of collecting high quality (usually regulatory generated) study data is that this data can be used to predict toxicities of unknown chemicals. For this to be effective, we must have a very high quality, reproducible test, usually following a very prescriptive protocol,” said Dr. Clive Roper, Head In Vitro Sciences at Charles River’s Edinburgh site. “So for any of your toxicity outcomes you can start to identify the structure of the chemical that is eliciting this toxic response. Once you generate huge amounts of this data, you can then start to data mine toxic alerts for new chemical entities.”

Game-changing technology

Data mining this hypothetical repository would be a game-changer. Researchers could enter the chemical structure they want to test and the computer program could predict whether the chemical has an alert for different toxic responses. Think of it like a giant, virtual lab rat – the chemical inputs and their subsequent effects are already there in the data.

“So it would actually be measuring the animal responses from all the different experiments that you put in there. From a 3Rs perspective, those animals keep on generating data for many years to come,” Dr. Roper said.

The science behind this has been thoroughly considered, and the results are promising from commercial software using peer review data. Adding the standardization of SEND to the equation makes this virtual animal model even more promising. However, there is a catch: companies would have to cooperate with competitors by sharing their data.

“Ultimately the blocker here is intellectual property – who owns it?” Dr. Roper said. “We think that anyone who signs up to have their data harvested, would be given access to test their NCEs. It would be dealt with by an IP lawyer who would say that, ‘If you sign up to this, then you have full access to what comes out.’”

IP issues a concern

It is a careful balancing act. On the one hand, the benefits for everyone are huge. Fewer animals means lower costs, and researchers would have the opportunity to throw a lot of chemical spaghetti against the toxicity testing wall without wasting any lab time or materials. On the other hand, companies who pay for expensive tests are understandably reluctant to see their data go toward helping a competitor. Each company will need to decide whether the benefits of cost savings outweigh the loss of data secrecy.

“Every time you put another data point in you’re going to be refining this model, becoming more accurate,” Dr. Roper said. “If you’ve got all the CRO’s tapping into it you’ve got a much, much more powerful tool. You maybe can’t see the input data results as this would be completely blocked off for confidentiality reasons. But the more you put into it, the more powerful the tool becomes.”

Dr. Roper presented on this topic in a September 2018 meeting of the Scientific Advisory Committee on Alternative Toxicological Methods (SACATM), part of the US National Institutes of Health (NIH). In his presentation he stressed that using Big Data methods would not only cut costs and reduce the time to market for new chemicals, but would also conform to the 3R’s of ethical animal testing: Replacement, Reduction and Refinement.

According to Dr. Roper, the tool should be chemical agnostic. That means any chemical product could be tested, whether it is a new drug, chemical, or agrochemical. As he says, the body responds to chemicals regardless of what we call them, and so will a virtual rat.

Our four-part series on predictive toxicology continues tomorrow with a look at regulatory patterns.