A huge thank you again for the incredible outpouring of support you have all shown. The project has already far exceeded any expectations we originally had of what we might be able to achieve! Last Thursday we closed the second round of submissions and wanted to share some important updates with you…
Submission Summary
We received a total of over 3500 submissions from 400 contributors across the two rounds. The summary of all submissions, duplicate designs, as well as other informative splits of the data can all be found in this post.
We have now sent the first 250 compounds to Enamine for synthesis – these form the ‘regular’ track as discussed here. These were selected almost completely based on synthetic feasibility and/or purchasability. About half of the compounds are contained in Enamine REAL space, while others were found with our technology to be easy 1-3 step syntheses from Enamine building blocks or readily purchasable compounds. Below you can view the full set of these compounds, along with the synthetic routes (for the custom compounds) that PostEra’s algorithms have generated. Within the route viewer there is the option to provide feedback/comments and we would most certainly appreciate your input to improve our process going forward.
The team is now selecting the ~15 compounds for the ‘fast’ track. These compounds will be selected as mentioned in the forum.
The Ask
We are aware that synthetic accessibility is by no means a perfect selection criteria – and we are now working to select more complex compounds from the submissions. If this was an official benchmarking competition, or chemical A/B test, we would work better to define selection metrics – however given the pressing demand of the situation, this is a bit more like grading a literature paper than a math test. We are in a pandemic, and our goal is rapidly work towards an antiviral, so we need to get moving and have things made! However, given our desire to not lose out on all the extremely valuable input the community has given, we are now asking the following of you: Pick your 10 favorite molecule designs by other submitters. (Also, please comment on the designs along the way!).
We will then compile these results and look for those that show up multiple times. The web page is now randomised for each viewer to make sure there is no bias toward the ‘top of the page’ designs. Your picks can be based on rationale, quality of merge, your own docking results, etc… just please submit your opinions in the following way:
To submit your picks please reply to this post with a .csv file in the following format:
Hi @JoostU, so the second round of submissions has technically just finished – we are now just looking for people to submit their favorite designs of others from all previous submissions. No decided deadline as of yet (perhaps end of week/weekend), but I’d say the earlier we get it – the more likely we are to use the suggestions.
And realized my title was a bit confusing, so changed it it. Thanks @JoostU!
first of all great initiative and great work! Awesome to see how the community puts together all these ideas!
Just having a few questions/comments:
I quickly went through the 250 selected compounds, did you consider using a diversity picker? I feel some of the compounds are very close to each other and we might have a higher chance finding something starting from a broader chemical space, no?
I’m wondering how you did the search for the Enamine IDs (especially the REAL space), not sure if you got all compounds (I just rechecked some of our submission out of curiosity, they are flagged as false, though they do appear in Enamine REAL space).
I may have missed some overview of the strategy/next steps:
Are you - besides the selection of fast track compounds - still interested in selecting more compounds from all current submissions Covalent and non-covalent ones? If so what analysis would be helpful for you that we, the community, could perform to prioritize compounds.
Are you generally interested in more submissions? If yes, it would be super helpful so define and announce the selection criteria for the compounds, that helps focusing on designs
I reviewed the 250 selected compounds and I agree with you. I understand the fact that we are facing a pressing demand due to the urgent need of moving forward science in this area. Because of this, we should also focus our attention and resources on robust, novelty focused, designs even more. We might really have a chance to select, as a community, compounds which are highly diverse, designed by experts and that can really be a gamechanger. I think this is our chance as scientific community to have impact on fighting COVID-19. We cannot waste it.
I have the feeling that the majority of the selected compounds are more fragment-like than lead like. If active, it might be needed to invest consistent amount of resources to increase protease affinity and generate molecules with better PK properties at the same time. This is not light work. The chemical modifications that might be needed to consider to improve compounds properties might also lead to important changes in the synthetic routes.
I think we should move forward with audacity and scientific rigor to generate transformative compounds with a good chance to being fast tracked towards clinical trials.
I am very happy to provide input, what do you think of organizing a virtual round table?
@mc-robinson will give a more detailed answer, what I can say quickly:
2. see this topic:
It is not so much what is potentially available as to what Enamine can actually make, in 2 weeks, from available building blocks. So there are further constraints.
3. Yes interested in both. Our current bet is that we have a better chance in gaining potency by focusing on covalent compounds for the “fast-track”. However we are using as many tools as we can to evaluate the quality of the designed compounds as possible, it is not straightforward. All the compounds designed are taken into account for the standard route.
4. Yes of course, the aim is to be an open data platform. The core team is working hard on trying to find the shortest way to impact, but the data is available for everybody to use, now or later.
They are very fair points and will be discussed in todays meeting. The compounds from the first call were fairly simple because we need a fast turnover for now and test every stage of the platform. Challenging syntheses require time, operational labs and money. What we are hoping is that instead of making the synthesis complex, a smart merge of compounds would provide the desired leap in potency.
Hi Anthony, thank you for your message.
I agree with you. A smart merge of compounds might lead to more potent compounds. I looked though the list of compounds which are progressed further in the pipeline and I compared them with the original fragment hits. I have noticed some very simple substitution patterns. For example a fluorine atom that was substituted to a methoxy group. Experienced scientists and free energy calculations expert will agree with me that this sort of modifications should be addressed in a lead-optimization phase rather than a hit-to-lead phase. The original fragments hits might not have the the right range of properties, including potency, to be progressed in higher tier screens. I believe this could be the case also for some of the compounds that have been progressed.
On the other hand we have some very exceptionally well conducted designs. It would be a pity to not value the solid efforts made by the scientific community as a whole.
We could also follow-up those and try to simplify the final design, to get a smaller, novel scaffold that is easier to synthesize and brought forward in the screening cascade. I am very happy to help.
First, thank you very much for this beautiful challenge. That is a very nice approach to contribute in croudsourcing manner in order to fight this tremendous outbreak from scientific perspective.
However, it seems that vast majority of selected molecules are too simple and non-covalent, while the main focus indicated recently was to design covalent ones as they are much more promising binders. I agree with davidesabbadin that the synthesis of ~200 simple and less promising cpds would be a waste of time, while the synthesis of 25-30 more complex and affine ones with higher scores in modelling would result in much more probable binding to the protease, even if it took 3-4 weeks instead of 1-2 weeks to synthesize them. IMHO the quality here should be a priority rather than quantity. Also I agree that the selected cpds are very similar while the pointed focus was to select diverse cpds. Another selection criteria was pointed as to select interesting cpds instead of simple Community input:If we and others are particularly impressed by a design or the rigor of a design methodology, we may prioritize those compounds despite more difficult syntheses. I believe there are many interesting examples worth to be synthesized and tested against the protease. Thank you !
I agree that the selected designs are quite fragment-like, and not very diverse. (And you had explicitly asked us to find ways to merge fragments into lead candidates!)
I agree that community involvement, ‘the ask’, is definitely necessary to scale your ability to select interesting candidates. I wonder if motivating with a little bit of self-interest some open peer review might help:
Allocate a random submission to each user to peer-review (comment on, and rank)
Each complete peer-review moves your submissions up the peer-review priority list (so more eyes see it)
I think the peer reviews could go into the discuss discussion post associated with each submission, which also gives a venue for the submitter to discuss.
I’m worried that if we just go voluntarily to find our favourites, a vast quantity of the submissions will never get critically appraised, and we may just be biased towards big names / labs, and trendy methods.
It could be interesting to open peer review the designs to a broader scientific community. It would be a great idea to have input from experienced drug designers from industry and leading academic institution and use their input to complement the enormous work achieved so far by the challenge organizers.
Quick response to this. This first set of compounds (250) is from the first call (2000 molecules designed) and what best suited accessibility on a 1-2 weeks leadtime. The list from the second call (focus on covalent) is still being finalised. Thanks for your feedback, we will take it into account!
As the main focus is on irreversible inhibitors and suggested molecules must contain the FBS cores revealed recently by X-Ray, the attached compounds are more attractive. They can be resonably regarded as good templates considering novelty issue and MCF scope.
Unfortunately, a majority of the uploaded structures to date are “trash” poisons at all.
Selected Compounds.csv (890 Bytes)
Provided compounds satisfy both of the selection requirements, such as covalent binding ability and design based on recently obtained X-ray structures. The proposed structures were obtained by state-of-the-art rational design, demonstrate favorable binding positions and possess drug-like properties.
Hi, wow glad to wake up to some good discussion. @AnthonyA has had a nice overview of some of the reasons we’ve ordered what we have ordered so far, but I’ll just add a few thoughts. (And try to answer most of @andrea, @davidesabbadin, @BogdanZag, @jarvist, and others concerns)
As @AnthonyA mentioned, these are just the very first compounds we have ordered. More are coming very soon. Yes, many are fragment like because that is what is quick and cheap to make within a week. We are still very interested in bigger merges and more complex designs, which will be ordered soon, but many of these are complex syntheses and triaging them takes a bit longer. Even one of the “almost perfect” merges in the fragments, which will very likely be made, is in the 6-10 step range and is not technically “ordered” yet. We are still figuring the whole “live tracking” thing out, so bear with us.
In general, I think that most of us are fully aware that a one step synthesis fragment is not going to be an amazing lead. However, if you simply go to complex molecules, others will simply complain that we are missing the whole point of fragment based design! So, how do balance the advantages of FBDD – and build in potency early from very small changes-- with our larger ambitions to make more complex molecules quickly. Our quasi-answer right now is to do both in parallel. We are just taking a bit longer to triage the complex syntheses.
And since those are more of an investment too, we are asking for everyone’s feedback: thus this thread!