Skip to main content

How machine learning can help tailor the future of medicine

Researchers in the Department of Physics are exploring the electrical workings of nerves and how they can make the drugs of tomorrow safer and more effective.

A close-up of white, lozenge shaped tablets.
Understanding ion-channels could lead to the development of safer and more effective drugs.

In the body, cells exchange electrical signals, which are carried by ionic charges. Large protein molecules called ion channels open and close, moving these electrically charged ions across the cell membrane. Found in ‘excitable’ cells like neurones and muscle cells, ion channels are essential to the signalling processes involved in muscle, heart and brain function.

They can be thought of as the transistors of cells, controlling a cell’s ability to send electrical signals normally. When these proteins develop defects and ion channels mutate, cells stop functioning in the way they should and can result in neurological conditions such as epilepsy, multiple sclerosis, and even brain cancer.

For many researchers exploring ion channel dysfunction, the goal is understanding how each drug affixes to ion channels and solves the problem by restoring the appropriate response. Each of the main families of ion channels, however - sodium, potassium, and calcium, for example – may have hundreds of subtypes. Tracking down the one that’s causing the disease poses quite a challenge.

Professor Alain Nogaret and his team in the Department of Physics are combining physics-based optimisation methods, electrophysiology and drug discovery to tackle this problem.

Locating dysfunctional ion channels

The currents that pass through a cell’s ion channels are very tiny, measuring just a few pico-amperes, making it almost impossible to monitor individual channels over time in search of mutations.

Working alongside cardiologists and physicians, Alain and his team are using non-invasive, computational methods to look for changes in ion channels.

In the same way that a skilled mechanic can listen to a car’s engine and detect a faulty part from the sound alone, the team can ‘listen’ to our cells. In this case, the sound of the engine is the electrical oscillation of the cell, which is easy to measure.

By applying a complex stimulation, they can extract as much information as possible from this cell ‘noise’. They can then use machine learning to disentangle the contribution of individual parts and see what current each ion channel is letting through the cell membrane at any given time.

Monitoring this cell ‘noise’ over time will allow the team to spot any changes in the movement, conductance or threshold of ion channels, all of which might indicate possible malfunctioning.

Applying the science

One of the biggest challenges in drug development is the inadvertent side effects of new therapies. A new painkiller, for example, could target key ion channels in the central nervous system, reducing the number or strength of pain signals being sent. With so many ion channel subvariants present in cells, however, the new painkiller could target channels it wasn’t meant to. If it were to interact with ion channels in the heart instead of just the nervous system, patients could suffer undesirable side effects such as arrhythmia, or even cardiac arrest.

A key advantage of this method is to monitor the activity of multiple ion channels simultaneously, probing alterations in the ion channel targetted by a new drug and also alterations eventually incurred by the other ion channels. Not only would this allow them to determine whether the drug will be successful, but it would also act as a drug toxicity screen and highlight any potential side effects during drug development.

Another important application could be in tailoring treatment to individual patients. Different people have different numbers of ion channels in their cells and can react very differently to the same drug.

Currently, many practitioners prescribe drugs on a trial-and-error basis, tweaking drugs and dosages until they see success. This approach is time-consuming, costly, and potentially involves risks to the patient.

Using a technique like the one the team are proposing would allow clinicians to analyse which ion channels are present and allow for personalised medicines and therapies for individual patients to be prescribed.

The future of pharmacy

In years to come, detecting ion channel dysfunction could act as an early marker of disease onset. The team is working with clinicians to investigate arrhythmic cardiac cells and ion channel dysfunctions in brain tumours.

Alain and his team are currently focussing instead on growing their network of partners and moving their research out of the lab and into the real world. They have recently signed a new memorandum of understanding with Masaryk University in the Czech Republic, working with cardiologists to access patient data for analysis.

They’re also collaborating closer to home, working with Dr Chris Bailey from the Department of Life Sciences. An electrophysiologist, Chris can take the recordings the team need to analyse ion channel activity. Previous data has come from partners in other universities across the world, however, Chris’s expertise in neurology and pain relief adds an extra dimension to this research.

Related case studies

Find out more about research in the Department of Physics.