Kenneth Baillie

But finding several therapeutic targets so quickly is extraordinary. There are immediate implications. I'll take one example. A DNA variant near TYK2 is more common in people who need intensive care for Covid.

Your DNA is a long code, which we represent as the letters A,C,T and G. There are 3,000,000,000 letters the code to make a human. At this one position, if you have a "T" instead of a "C", then your odds of life-threatening Covid are 1.3x greater

Doesn't sound like much, and compared to the effect of age on risk, it isn't. But that's not why it matters. That one change makes a difference to how much of the TYK2 gene you make. So we can ask, if you make more TYK2, are you more at risk?

The answer is yes. Less TYK2 is associated with lower risk That suggests that a drug that inhibits TYK2 might make people less likely to develop life-threatening Covid. The good news is that we have a whole class of drugs that do this (JAK inhibitors).

The other genes we find suggest other treatments, which we discuss in the paper. We already know that genetic evidence doubles the chance that a drug will be successful.

This demonstrates the beauty of genetics for drug target discovery. Faced with a new disease, that we didn't understand at all, we can look across the *entire* code that makes our immune system, to find the exact points we need to target with drugs, in order to save lives.

The funding that made the @GenomiccStudy possible comes from the relatives of patients who died of sepsis (@stopsepsisnow), and we're trying to do the same thing to find new treatments. I told them it could take 10 years to see a signal, and even then it might not lead to a drug

It's astonishing that we were able to find therapeutically-relevant genetic discoveries within 6 months of the first ICU cases in the UK

Of course, as we say in the paper, we can't know that any of these drug predictions will actually work until we do large-scale, randomised clinical trials, such as RECOVERY

More than 1400 research staff worked on this project. We can't thank them all but @erolaPC @thefluzee @kcrawlik @l_klaric and social media hold-out Andy Bretherick worked day and night for 6 weeks to get the results out. @LeeMurphyCRF @EdinburghCRF did all lab work

@GenomiccStudy team, led by Fiona Griffiths, have suffered all year to cope with the massive surge in cases. 64 volunteers from @TheDickVet and @roslininstitute came forward to help when they were struggling.

And 2244 patients, or their relatives, agreed to contribute to a research study to help others, at one of the most difficult times in their lives.

The work was funded by @stopsepsisnow @wellcometrust @ICS_updates @UKRI_News @DHSCgovuk and carried out in partnership with @CCPUKstudy and @GenomicsEngland

Thanks for making it this far into a long thread. We now have 6685 patients in GenOMICC. But the best way to study them is to find similar people (by age/sex/ethnicity/postcode) to compare them to. If you'd be willing to help, please go to http://genomicc.org  and register

This thread is aimed at the lay public - we already reported our findings, and shared all our data, with the genetics community on 27th September. So I've glossed over some nuances which we discuss in the paper.

The biggest limitation is that the genome is complicated and our predictions are based on the assumption that there is only one mechanism linking the genetic variant to the clinical consequences. That's why I keep saying no-one should treat patients on thia evidence. Need trials.

The key point is that this is big progress, but not an answer. If one third of our drug predictions from genetics lead to effective drugs, that's much better than the current success rate in trials in covid. And infinitely (literally) better than the success rate in sepsis.