Clinical trial is tricky. On one hand, some researchers would risk everything from the fabrication of data to publishing falsified results, to push forward their innovations. On the other hand, patients misbelieve clinical trial as their new hope. Some would share their trial results on online support groups and jump from one trial to another, expecting to find the quickest cure. In either scenario, data plays a pivotal role. Good, actionable clinical data is required to generate a new medical solution. There is also a need for a representative sample to try out whether this new medical solution would work. 

In silico medicine  

To perform clinical trials in virtual bodies or In silico medicine, the use of computer-based technologies – systems, algorithms, data – to model, simulate or visualize biological and medical processes, was one of the proposed solutions. HumMod, an initiative by researchers from the University of Mississippi Medical Center, has begun building a complex mathematical model of human physiology decades ago. The model has successfully given rise to several projects. 

A few years ago, Harvard University’s Wyss Institute reciterated the microarchitectures and functions of living human organs within computer microchips. Known as “Organs on Chips”, the creation renders an alternative to animal testing as major body parts such as lung, kidney, skin, intestine, blood-brain barrier and so on, could be replicated artificially. 

In Europe, the In Silico Oncology Group had developed an experimental platform and an advanced medical decision support tool called Oncosimulator to improve cancer treatment. The Oncosimulator is an integrated software system whereby it virtually mimics the responses of a tumor to therapeutics in a clinical trial setting. 

With artificial intelligence (AI), some of these simulations, especially those concerning drug development, could answer question such as “why would this procedure work for this patient but not other”. Likewise, neural networks could effectively predict if the target medication or medical procedure will act in the best interests of a patient. 

Are we solving the fundamentals? 

In terms of regulation, the US’ Food and Drug Administration (FDA) had already planned for a possible future whereby over half of the clinical trials data are coming from computer simulations. However, a commentary, co-written by experts from Harvard School of Public Health, Johns Hopkins University, Massachusetts Institute of Technology, New York University, University of Toronto, and Microsoft Research, published in The Lancet at the beginning of August, questioned whether we can fully trust algorithms. 

Primarily, clinical data are still in a mess and there is not sufficient actionable data to train and test an algorithm, let alone building virtual human bodies that best represent a comprehensive group of patients. At the same time, it is hard to deduce if the new drug or medical procedure that had passed the virtual clinical trial, could account for any possible unknown. Ultimately, who is going to be liable, shall the new drug or medical procedure suffer some wrongdoings which endanger patients? 

Most importantly, both AI and in silico medicine are still at their infancy. Whilst they show potential in solving existing problems, it remain unclear if new problems may emerge. 

Author Bio
synthetic gene empathy chinese artificial intelligence data medicine healthcare ai

Hazel Tang

A science writer with data background and an interest in the current affair, culture, and arts; a no-med from an (almost) all-med family. Follow on Twitter.