Clinical trials, especially human trials, have come in for some stick in recent years, with some disastrous results that have threatened people’s lives. The TGN 1421 test that left several test subjects with long term damage was widely publicised after it occurred in 2006, which saw several changes to the laws surrounding drug testing and the patients receiving millions of dollars in compensation. The changes that this case brought about have undoubtedly improved drug testing, but recent big data and machine learning developments have pushed both safety and speed to a new level.
Weill Cornell Medicine scientists have created a machine learning system that can detect the level of harm that a drug could potentially have to a human. Dr. Olivier Elemento, who helped develop the technology, said: ‘We looked more broadly at drug molecule features that drug developers thought were unimportant in predicting drug safety in the past. Then we let the data speak for itself.’ Essentially, through looking at the results of thousands of drug trials, the system could pick up molecular elements that caused specific toxic reactions in previous drugs.
The new method is known as PrOCTOR, with the idea being that the system will be able to give ‘PrOCTOR scores’ to new drugs, allowing companies to have faith that they will not have harmful side effects.
This move is not only going to be useful for increasing the safety of drug trials, though, it is also going to help companies increase the speeds in which drugs are brought to market. Today they need to conduct extensive tests beforehand, just to establish whether a drug will have harmful effects, which is both time consuming and expensive. With this new system they can use predictive modelling to understand which drugs are going to cause harm and which aren’t, without ever necessarily needing to make a single pill or vial.
It is this move which may well have the biggest single impact on the development of new drugs, which could have profound impacts on fighting diseases that are quickly spreading. For instance, the Zika virus has been growing in prevalence with several children born with abnormalities because of it. However, it doesn’t have any kind of vaccine and the time it takes to formulate, test and produce one had put thousands more people at risk. Through utilizing predictive modelling in this way, pharma companies can do at least the early parts of testing considerably quicker, potentially developing these essential drugs much more quickly.
Many drug trials don’t actually test new drugs but instead test drug combinations to test how many varieties work together. These can often be the most dangerous as there are multiple different elements to consider, with thousands of potential side effects. Through modelling, it is possible to identify what these may be, with considerably better accuracy and speed than with traditional methods.
We have seen data have a profound impact across many parts of our society, but this latest development may well rank as one of the most positive for society.