Finding phenomic fingerprints of COVID-19
One of the biggest challenges posed by the pandemic is telling who is most at risk. Eighty percent of people will have very few symptoms, and maybe half of those may even have no symptoms at all. But one person in five will develop more severe - and in some cases life-threatening - disease. Now, a project set up between the University of Cambridge and another university, literally on the opposite side of the Earth, may have a solution: a chemical fingerprint present in the bloodstream that predicts whether someone is destined to get a very mild form, or a more severe form of the disease. And this matters, because knowing in advance who is likely to have trouble with Covid means doctors can intervene earlier with drug... and other therapies, potentially preventing severe deterioration. Chris Smith spoke to Jeremy Nicholson, From Murdoch University in Perth, Western Australia…
Jeremy - The challenge that we've set ourselves, is to use a range of advanced technology platforms for very rapidly profiling the chemistry of blood, and possibly urine samples, from COVID patients to try and predict how severe their disease will become. So at the moment, we have a number of different ways of testing for the presence of the virus, and potentially how you immunologically react to the virus. But what we can't say in the early stages is are you going to need to be on critical care or not? And if we could detect a signature for that, then we'd be able to manage those patients much more efficiently.
Chris - What might that signature look like Jeremy? And how do you go about finding it?
Jeremy - We use a range of advanced technologies you normally find in a chemistry laboratory, they measure hundreds or thousands of different molecules that are in your blood or in your urine. So the first real question is: can we find out whether or not, using this range of different chemical technologies, that COVID patients are different to everybody else? And what I can tell you is that our initial tests show that that probably is the case, but what really we want to know is whether or not that signature predicts the severity going forward. For that, what we need is longitudinal sampling. So we have a series of different samples taken from a bunch of patients, and we follow them through and we build a model of how their chemistry is changing with time. At the end of the patient journey, we will know who is severe. And what we're going to do is look at the earlier stages of the chemistry to see if there was something very specific associated with that turn towards severity.
Chris - So it's not just one molecule that changes, you're talking about a whole constellation of molecules, and how the levels of those different chemicals relate to each other. And there might be a particular spectrum, which is very specific to people who are going to get severe disease, compared with people who are not. And if that's present before they even get infected even, or very early in the course of the disease, you can say those are the people that we need to intervene early, or they're the ones to watch, we could reassure the rest.
Jeremy - Well, that's absolutely right. I mean, we measure probably as many as 20,000, 25,000 compounds using these different technologies. That's lots of different types of small molecules, but the signatures come down to a much smaller numbered subset of those. Maybe only 10 to 20 of those things will be really reporting information that's relevant, that's specific and sensitive for the disease and also, reporting on how the disease is likely to progress.
Chris - How are you going to find them in the first place though? Because this to me is, you're hunting for a small number of needles in a massive biochemical haystack. So how do you know what those needles are in the first place?
Jeremy - Well, the way that you make the predictive model, is that you have to have samples that you know are from people who, shall we say only, had the mild disease and others, which went on to get the severe disease. And then you do a statistical analysis to find out which of all the different combinations of metabolites in there are the ones that are most closely associated with the biological question, which in this case is the prediction of severity?
Chris - Where are the samples going to come from that you're going to analyse? Is this just Australia?
Jeremy - Potentially samples from several places in the world, but one of them, of course, is your own institution, Cambridge university., Where Cambridge, I think the Addenbrooke's hospital has handled more samples than have probably gone through the whole of Australia. So there is some very well designed studies there on COVID positive patients at different stages of the patient journey, and with different degrees of severity. And we are in the process of having an arrangement to ship some of those samples over and do a comprehensive analysis.
Chris - And just lastly, some groups have stood out as being potentially more vulnerable than others. I'm thinking older people, people with male sex, also people from certain ethnic backgrounds. We don't know though, to what extent socioeconomic factors is playing a role. So is this going to help us to disentangle that aspect as well?
Jeremy - Well, all of that background information would be important metadata that we would co-analyse. And of course there are lots of different reasons why all of those things could contribute to your risk for, not just for COVID-19, but for lots of other diseases as well. The important thing is though, you're measuring the contribution from genes, and environment, and diet, and all your lifestyle into your individual metabolic signature. So when we run our various fancy machines, we are picking up the signature of your life. And the signature of your life also, is very closely related to your disease risks for anything. And that includes severity of disease. So the nature of the science that we do, unraveling the gene environment interactions that creates the metabolic phenotype, which is related to the risk, by definition, it will be related to those other risk factors, as well as age and gender and ethnicity