A virtual model of a human: science fiction or future? How do we do in silico research?

Rustam Gilfanov, a Venture Partner at the LongeVC Fund shares his ideas.

Science and medicine develop today at great strides. But is it possible to create a computer model of a human body? Is this something the future holds for us?

In the very beginning of the confluence of medicine and IT technologies, computers were regarded only as an instrument to facilitate the processing of large masses of data. Now science community is thinking of (and working on) holding fully virtual research experiments. A virtual model of a human body would allow medics to create new drugs easier as well as make personalized drugs more available.

How are in silico experiments held?

The first mention of in silico (literally “in silicon”) experiments dates back to the 1990s. Today this term is used almost as often as in vivo (“within the living”, i.e., experiments on whole living organisms) and in vitro (“in the glass”, i.e., experiments held in a test-tube artificial environment).

By the in silico experiments we mean putting certain parameters in the computer in order to recreate the results that would be obtained during a real experiment on an analyzed system. At present in silico research also includes predicting the behaviors of certain molecules, biochemical processes, and full physiological systems.

Special attention should be given to in silico modelling of individual molecules. In this case, a computer models the reaction of various molecular systems, for example, amino acid or nucleotide sequences. Not to be forgotten that computers are now indispensable to studying big masses of data (bioinformatics), just as they were in the beginning.

Empirical force fields-based studies are yet another application. They model the spatial structure of protein molecules depending on structural templates of proteins with related amino acid sequences. A computer can also study interactions like in pairs ligand-receptor and enzyme-substrate — this aspect is widely used in designing new drugs, and it is related to as studies of molecular docking.

So far, we are not at the point where we are able to use in silico research instead of other forms of human experiments. However, it is an auspicious direction for development. The most obvious reason is that in silico models can save time and money spent on running in vitro and in vivo experiments.

Let me give you an example. Within a de novo drug design a computer can design a molecule that would perfectly connect with a necessary part of a target molecule that should be bonded with a drug molecule with which the drug needs to form bonds. Also, it can model the behavior of a new substance by analyzing its chemical structure in order to, for instance, predict the toxicity of reactive impurities.

Currently, the majority of in silico models rely on existing databases of substances and their toxicity. The PMI company already uses in silico for their research on tobacco. Recently they introduced INTERVALS, an open online platform accessible to other researchers. It gives access to the results of completed tests and protocols, thus saving time on checking new hypotheses and holding research. It is supposed to make big data analysis simpler as well as help schedule new tests. 

Digital twin startups

Another promising direction of in silico research is the creation of virtual models of individual patients. It gives scientists an opportunity to hold tests with rare phenotypes that are difficult to find for real-life trials. Also, this approach allows for comparing the efficacy of various treatments in a certain patient without spending too much time.

Recently a joint UK and the Netherlands research team under the management of Alejandro F. Frangi (the University of Leeds) successfully proved that a virtual model offered exactly the same results that conventional trials with real patients. The model in question analyzed the treatment of brain aneurysms with a flow-diverting stent applied to a sample of virtual anatomies available in clinical databases. Put more simply, a stent is a tube placed in the artery in order to reduce the blood flow, thus lowering the risks of rupture and ischemic stroke.

For this experiment scientists selected 82 virtual patients with similar age, sex, nationality, physiology, anatomy, and biochemical characteristics to those of real people who were taking part in a trial of the stent efficiency. For these 82 virtual patients, they developed a program to analyze the influence of the stent on blood flow and then compared their results to those of three real medical trials.

According to the virtual model, the use of the stent was “virtually successful” in 82.9%. The success rate in real trials used for comparison was 86.8%, 74.8%, and 76.8%.

Another good example of the technology’s application is HostSim. It is a virtual model of a host of Mycobacterium tuberculosis developed by a group of scientists of the University of Michigan Medical School. This model analyses the response of the host’s immune system to the pathogen and predicts how the disease is going to be progressing. A virtual model was used because it is very hard to collect samples from the infected lung granulomas of a patient. This virtual primate model had lungs, LN, and blood: the systems that are most affected by the progressing disease.

Another two projects — The Living Brain and The Living Heart — were created by Dassault Systemes (France) on the 3DExperience platform. After studying available information from patients and various studies, the scientists developed a virtual model of blood circulation that allows them to study different treatments. With the help of the Living Heart, they can create new medical gadgets, test the safety of drugs, and develop personalized surgical treatments.

The other project, Living Brain, helps study epilepsy and determine which parts of the brain are associated with seizures. By 2019 the results of this project were so good that the FDA prolongated its work with Dassault for another five years.

In Germany, the Ebenbuild company developed a personalized virtual twin therapy for ARDS patients (Acute Respiratory Distress Syndrome). In order to create a virtual twin, scientists use the patient’s data from CT scans and process it using AI and image analysis. As a result, they get patient-specific lung segmentations that with the application of lung ventilation parameters are used to increase the patient’s chances of survival and recovery. 

“Local mechanical overload of the lungs due to suboptimal ventilator settings is a major contributor to the high mortality in patients suffering from Acute Respiratory Distress Syndrome (ARDS)”, state Ebenbuild scientists on the official website. “Our technology enables us to provide the best possible protective ventilation protocol for each patient, reducing ventilator-inflicted lung damage. Combining a CT scan of the patient’s lungs with in-depth physiological knowledge, engineering, and physics-based algorithms, we create highly accurate digital twins of the human lungs”. 

Could we create a full human digital twin?

The rating of popular digital technologies in the sector of healthcare LIFT Radar 2021 put a human digital twin on second place and the IEEE Computer Society rating of 2020 ranked it third. An analytical leader Gartner in 2020 declared that digital human models will be a highly valued technology from a social, healthcare and business point of view within the next 10 years.

So far virtual twin technology is only in its infancy. Some of its approaches are used for the creation of new drugs and certain types of personalized therapies. However, the same approaches are already widely applied in other spheres — in the construction, manufacturing, and automobile and aerospace industries. Engineers use them when designing new systems, managing exploitation and calculating possible equipment wear in real time.

It is impossible to create a model of a human patient without a constant update of genomics, biomics, proteomics, and metabolomics data, as well as physical markers, demographic, and lifestyle with regard to the timeline . However, as medics have more and more devices for data collection, the enthusiasm for the digital twin grows accordingly.

Ideally, all this data should be taken in real time from the real patient and delivered to his virtual twin. This way the medics would have the most up-to-date status of the health of the patient and be able to find the best treatment accordingly.

Chinese scientists Lui, Zhang, Zhou, together with their colleagues in 2019 formulated the conditions for the creation of a virtual human twin. First, some kind of advanced modeling tool (SysML, Modelica, SolidWorks, 3DMAX, or AutoCAD) should be used. Second, the healthcare Internet of Things (IoT) and mobile connection should be employed for data feed in real time. Third, a model should be subject to routine calibrations. And the last one, the patient should receive the results of any activity applied to his twin model (e.g., diagnostics).

What is now obvious is that the creation of a virtual twin technology would undeniably lead to a medical revolution of our age. Personalized treatments, prevention of diseases and use of patient-specific therapeutic methods based on genetic, biological, phenotypical, physical, and psychosocial peculiarities would become a new medical breakthrough.

Weaknesses of the in silico technology

Currently, the level of technological advancement does not allow us to predict all possible influences of a newly created drug, so for the moment, computer-based research cannot completely take over real-life clinical trials. Thus, unfortunately, the amount of animal tests and time and investment consumption stays the same. Part of the problem is also the lack of trust to computer tests compared to more traditional approaches.

Yes, technological optimists stay very hopeful about the in silico research, but realists consider it very improbable that the technology would ever be able to model all the processes in a living cell. That would literally mean modelling life itself. For many people, this pragmatic approach to life is still something of the blasphemy towards nature.

Let’s assume that one day our technological level would allow us to decipher all the enigmas of the human body. Even with that, we would still need much more progressive methods of data collection and storage, not to mention the cost of medical devices. Until those are available to a wide market, no advancements can be made.

Another source of worry is data quality. A working computer model needs reliable data, however, at present many databases are not objective enough to be taken into account. For instance, take racial and gender bias: data available for white men strongly dominates in medical research.

So, the first problem to tackle is data collection and storage. For reliable automatic data collection and analysis, we need a large amount of detailed data and unified electronic health records. So far, electronic health records are dissimilar and unstructured, and for the most part, cannot be used for reliable computer analysis. There is also an ethical problem: doctors need full patient consent for data collection and manipulation, and this is where we face the issue of confidentiality.

Last but not least: with all the seeming complexity of a digital twin system it must be user-friendly enough to be used by the medics, the patients, and the computer, and ensure good communication between all three sides.

I would also like to mention an ethical problem of eugenics that would definitely arise with the introduction of virtual twins. Once computer models can reliably point out genetic profiles with high survivability, it would hoist old ideas of “good” and “bad” genes. From here there is only a tiny step to choosing embryos in regard to their genetic profile, and a little bit bigger step to checking the genetic codes of people applying for a job.

Yet another issue is people’s skepticism towards computer-generated solutions, both from medics and patients. There were separate studies of this problem that revealed the following: doctors tend to not trust AI programs introduced in the medical environment. Their main worries are misdiagnoses, wrongly chosen treatment approaches and, not to neglect, their concern that an AI would in the long run replace medical professionals.

About the author

Written by Rustam Gilfanov, an investor, philanthropist, and a venture partner of the LongeVC fund.

Leave a Reply

Your email address will not be published.