The end of the world was yesterday

The digital patient

When your avatar goes to the doctor and every diagnosis relies on a global database. How computers will make us healthier.

Hans Lehrach

In medical surgery, the personalisation of medicine is almost perfect. Imaging techniques generate millions of data points, allowing the surgeon to directly visualise this detailed information so as to adapt treatment decisions on a case-by-case basis. Implementing a similarly personalised therapy for drug treatment is much less straightforward: the doctor cannot directly observe the exact state of the complex molecular networks in the same way that the precise position of a fracture in a patient’s broken arm can be pinpointed. Moreover, it is even harder to foresee how these networks will be affected by the similarly complex actions of the drugs under consideration.

Today, drug therapy is therefore selected on a statistical basis. Your doctor will prescribe a drug that has performed best in a clinical trial for most patients with a seemingly similar disease. But this is not necessarily the drug that is best for you. Such an approach results in only a quarter to half of patients being responsive to the drugs they receive. Others show significant side effects, or may even die.

This problem is particularly dramatic in oncology, an area with on average very low drug response rates, frequently very serious treatment side effects, high drug costs, and sometimes quite limited survival benefits. Every tumour is unique – it has never been seen or treated before, and it will never be seen and treated again. Every tumour (and potentially every tumour cell) can therefore be considered a separate “orphan disease”. In order to extend the range of truly personalised treatments, we need sufficient clinical and molecular information about every individual patient. We also need a mechanism that uses this data to predict the effects (and side effects) of therapies.

Biomolecular research, systems biology and computer technology will provide us with tools that will enable medical doctors to find the best possible drug therapy for each individual.

The molecular key for future therapies

In recent years we have already made enormous progress. It took the Human Genome Project ten years and 3.6 billion US dollars to sequence the first human genome. We now have the technology to sequence an individual human genome within a few hours for less than 1,000 dollars. These new sequencing techniques can also be used to analyse specific tissues in more detail, providing information about a patient’s epigenome and transcriptome – i.e. genomic information that is “read”, processed and used in specific tissues.

These developing technologies will provide us with a more precise understanding of the processes underlying disease and our patients’ response to therapies. They enable us to analyse genomes or transcriptomes of individual cells and to characterise the status of the immune system in each individual. Huge progress has also been made in analysing proteins and the way they are chemically modified as part of the transfer of information within each cell. Similarly, we can also study metabolites, small chemical compounds with a range of functions including supplying energy, transmitting signals between different tissues and providing the building blocks for all the other, more complex components of our bodies. But we continue to strive for even better comprehension of our bodies in health and disease.

Biomolecular research, systems biology and computer technology will provide us with tools that will enable medical doctors to find the best possible drug therapy for each individual.

One strategy is “stratified medicine”. This approach uses statistical correlations between specific biomarkers (e.g. changes in the sequences of specific genes, the expression of these genes, proteins or protein modification states) and the response of patients to a specific drug therapy in a clinical trial. In most cases, however, this correlation is far from perfect, with no guarantees that the identification of particular biomarkers in a patient will lead to the expected response to treatment. Moreover, even the relatively few available biomarkers for stratifying patients with a single disease generate a huge number of combinations, making it technically impossible to test the full range in clinical trials. The combination of the results of even small numbers of biomarkers therefore very quickly becomes guesswork. Truly personalised treatment is therefore only feasible if we use a different approach: computational medicine.

Virtual bio-modelling: a new paradigm in medicine

This approach uses a detailed description of the situation at the beginning of the modelling of a patient’s disease and comprises a fairly detailed knowledge of its basic rules as well as possible disturbances, e.g. the mechanisms of drug action. If we want to build a new model of a car, we will first test thousands of potential designs in virtual crash tests, generating a great deal of information about its behaviour at much lower cost and in a much shorter time than in reality — and, most importantly, without endangering people. We now have the technology to apply this strategy in medicine to save many more lives.

Given our rapidly increasing capacity to characterise every patient in detail, the continuing rapid drop in computing costs and the ongoing investment in basic research that aims at identifying biological mechanisms, there is little doubt that this approach will play a key role in the coming years in replacing the current paradigm in medicine. Truly personalised therapy choices will be within reach, at first in areas such as oncology, where we have a rich and growing knowledge base but not yet the means of integrating all this data for individualised therapy.

Most, if not all, areas of medicine, prevention and wellness will follow. This paradigm change will be driven not only by the current revolution in molecular analysis techniques but also by the increasing power of sensors and apps that allow us to monitor our bodies and our biology in increasing detail.

Virtual twins

In the future, we will be able to rely on a “virtual twin” that is initiated at birth with our genome, continuously updated with information from medical check-ups or other treatments and brought together with an increasing amount of sensor data. This twin will be a “guardian angel” who warns us of dangers and the unintended consequences of action by ourselves or others. Such a “virtual twin”, who is only accessible to us, can be made available as an assistant to the doctor who is trying to develop an optimal therapy for a disease we might have. This would allow, but not force us to react more intelligently to many of the unavoidable challenges in our lives.

The same technologies could also help to increase the number of drugs available. On the one hand, any drug with known mechanisms of action can first be tested on a group of virtual patients. This would decrease development times, costs and risks. We could also scale up this scenario in virtual clinical trials, using the available molecular information that is currently being generated on a large scale by a number of global initiatives such as the International Cancer Genome Consortium and the Personal Genomes Projects.

In the future, computational modelling will be conducted as a prerequisite for clinical trials in order to determine whether the trial design – applied to a random patient population, potentially to millions of people with the same disease – has any chance of success. Of even greater benefit, however, would be the replacement of those large-scale clinical trials with much smaller ones on individual patients on the basis of biomarkers identified in the course of the modelling procedure. This change in strategy would not only help to reduce risks to patients but would also dramatically cut the costs, time and risk associated with the development of new drugs. Currently, only a few percent of drug candidates enter clinical trials, and even fewer end up being approved for treating patients. Virtual bio-modelling could increase this figure to 50% or higher, since drugs without toxicity problems could be approved for specific target populations, and even for small patient populations that are currently without appropriate treatments due to the high cost of the development process.

Challenges and prospects

In tandem with virtual patient technologies, we can also expect significant progress in other areas of medicine. Induced pluripotent stem (iPS) cell technology, for example, helps us to transform adult cells into something resembling an embryonic stem cell. These “embryonic” cells can in turn be developed into specific cell types, thus opening up new avenues for repair strategies based on cells that cannot be distinguished from the patient’s. One day we will be able to increasingly replace cells, e.g. the destroyed Langerhans islet cells causing type I diabetes, or photoreceptor/ AMD cells in degenerative eye diseases, or even organs such as teeth, using cells derived from other parts of a patient’s body, such as the skin.

Technological advances inevitably bring progress but also risks. Heated discussions are currently ongoing regarding the potential to engineer changes to our genomes that will be passed on to future generations. This will involve a very high level of risk, with potentially catastrophic dangers for the children yet to be born. In addition, it is really questionable whether humanity has demonstrated enough wisdom to use such a tool which has irreversible consequences. However, there are a few cases in which the medical benefits for future generations could at some point potentially outweigh the risks of such genome editing – for example, when the elimination of the receptor for HIV or other lethal pathogens becomes possible. Again, however, one must question whether other less risky approaches could achieve these goals as well, such as the successful development of vaccines.

For many applications for which we see potential technical routes, we still lack the requisite biological comprehension. However, future generations that have gained more knowledge (and hopefully wisdom) might judge our medicine more or less as we now view the medicine of the Middle Ages – as a crude and inefficient approach to dealing with nature. We are facing a new dawn of a more intelligent, data-driven and model-driven medicine that will without a doubt benefit all of us in the future.

About the author: Hans Lehrach is a former Director of the Max Planck Institute for Molecular Genetics. The Austrian chemist set up the company Alacris Theranostics, where he is a member of the Advisory Board, and is Director of the Dahlem Centre for Genome Research and Medical Systems Biology.