0:00
Thanks everybody.
0:00
So I'm going to be talking about is a little bit different topics from the last couple of talks we talking about resolving some post design bottlenecks in Med Chem workflows using some integrated tracking in tracking tools that we've been working on at Chemaxon.
0:23
Now we are part of Certara as of one year ago.
0:29
So I want to start by just mentioning the DMTA cycle.
0:33
I don't know if this is cliche yet or if everybody is tired of hearing about the DMTA cycle, but I think it's still like a really important formalism for the way that we approach chemical discovery.
0:47
This didn't exist when I first like got out of grad school.
0:53
I'm sure people were doing it, but we didn't think about it this way, but having, but I think having this framework really allows us to approach some of the problems that we have when we're doing chemical discovery.
1:06
Now, first, I'm sort of ignoring the fact that really what we're doing at the beginning of this is we're defining them like what we want out.
1:13
And as we go around the cycle over and over again, we want to be comparing what are we hoping to achieve with what are we actually achieving?
1:22
And are we actually getting closer to a target product profile for example.
1:29
But still, I think it's important.
1:31
It's important to think about this loop here.
1:35
And in particular, because kind of the longer we go, it seems like the people that are doing these different phases.
1:45
Maybe back in the day, the person that was making the compounds was also analysing and designing.
1:51
So maybe they had a bigger role and maybe they were sending things out for testing.
1:55
But more and more these pieces of the puzzle are being divided into two different people, different groups.
2:06
And that can lead to some challenges with how we actually go about this process.
2:12
I've highlighted just a few things here that used to frustrate me, you know, including like having data in different places and then getting data between each of these phases always seem to be a friction.
2:30
Of course, resources are limited, especially in competitive environments.
2:35
You know, more and more you have communication issues between internal and external teams and making sure everybody's on the same page.
2:44
Things that I always struggled with leadership in particular, where we're like managing multiple threads of ideas and compounds and compound sets and keeping track of all these things.
2:57
And then also keeping track of all those things in terms of like what's the big picture here?
3:03
And there were no tools like maybe 10-15 years ago when this was really like what I thought about all the time.
3:13
I mentioned that I was in the CRO game for a long time and there were no tools to help you along.
3:19
And now, thankfully there are several tools that can help us along, but in order to like, you know, address like how are we going to address those problems or how are we going to speed up the discovery process?
3:35
There's kind of two like maybe general ideas.
3:37
And this is just adapting a slide that I had at least 10 years ago when I was working in working the CRO space.
3:48
And we're trying to go around the cycle faster and prove that maybe automation is going to be the answer to some of these issues of like high cost of drug discovery and development.
4:03
And the challenge was, how do we know if we're going faster or how do we know that we're doing better or how are we going to do better?
4:12
And there's kind of two approaches, like are we going to maybe, you know, one thing we want to do is go around the cycle fewer times.
4:17
We'll make better compounds, we'll take advantage of AI.
4:21
Maybe we're getting close now to take being, really being able to take advantage of AI, stop chasing compounds that don't move toward the goal.
4:30
Don't remake compounds that you've already looked at in the past.
4:35
Let's do that.
4:36
Let's do this thing the fewest times possible.
4:39
And that's probably like maybe the more impactful approach.
4:42
But you know, as somebody that is trained in synthetic chemistry and was a person that was like managing synthetic chemistry processes, that's nice.
4:51
But also I think there's improvements that we can make ourselves and how we do things in the lab and at the bench.
4:59
So how are we going to go around this cycle faster?
5:03
Well, one thing we can do is like we can just try to make compounds faster.
5:08
One, the obvious approach is throw more chemists at the problem.
5:12
That's the expensive way to do things.
5:15
But often that is the answer.
5:19
Another thing that we often do is we restrict the chemical space we just heard about like building out 30 however… whatever unit that was chemicals, chemical space.
5:33
When you go to the lab, sometimes you start to restrict your chemical space in order to make compounds that maybe are not as good, but because the chemistry is easier, you can get it done faster.
5:47
And then there's also issues of like communication and hopefully we can streamline, smooth out hand offs between transition points.
5:58
So like, you know, these are some of the approaches that we might take.
6:01
And you know, again, as a synthetic chemist, I'm really thinking a lot about like what can we do in the make phase here?
6:10
And of course, like one thing about being in the make phase compared to all of these other phases is it's probably the one that's the most variable.
6:19
So if I send out material, if I send out materials for bio assay, hopefully the assays are standardised and I know I'm going to get my results back in a couple days.
6:31
Hopefully the analysis isn't going to last.
6:34
You know, maybe it's a day or five days or 10 days, but it's probably not going to be like a year.
6:41
And similarly with designing new a new set of compounds, but that make phase can be anywhere from, you know, a day to if it's a really important compound, maybe a year.
6:53
So thinking about like, how do we keep track just time that we're spending in the make phase?
7:05
So again, I mentioned that when I was really doing this myself, there weren't tools to help you along with that.
7:13
And now there are.
7:15
So our tool is called Design Hub.
7:17
We're happy to show it to you.
7:18
We have a booth upstairs.
7:19
We're happy to talk to you about it, but when I first saw this when I joined Chemaxon, I got really excited and I thought wow, why didn't I have this five years ago?
7:29
Because this is exactly what the kind of thing that's going to make improvements to our to the DMTA cycle.
7:37
So this allows us to have a hypothesis for how we're going to approach a problem.
7:44
Design compounds, keep those together in a logical way and like kind of keep our ideas together and then also help with the design and keeping data in a sensible place.
7:54
And it also helps with communication.
7:55
So it addresses a lot of those things that I mentioned a couple of slides ago.
8:01
But then there was still this like sort of lingering question about like how long is it taking to do things?
8:07
This is all good for designing and tracking ideas.
8:16
But then maybe so we need some project management on top of it.
8:20
And there's lots of project management tools out there now nowadays too, But there are very few that like would work well for chemistry just because you need chemical information as part of it too.
8:31
So built into Design Hub is this Kanban board.
8:36
And I think some of the other tools also have similar features.
8:41
But the Kanban board is great for having and sort of at a glance understanding of where we are on a project.
8:56
So I've simplified this to kind of just a handful of phases that you might have in terms of your, in terms of your workflow.
9:02
Maybe you have some compounds that are ready for synthesis, some that are in synthesis, some that are completed and cancelled that are maybe on hold.
9:08
You decided that we don't want to work on those and so we can just move compounds along down the track and we can organise those compounds based on who they're assigned to, how important are they?
9:22
Some other tags that we might, you know, some other properties that we might want to organise our thoughts about.
9:32
And so this is very nice, you know, for a project leader, you can come in, you can see where everything stands.
9:38
You can report out to project teams or leadership or whoever you need to talk to and you know exactly where you are.
9:47
And in particular, when you have diverse teams that maybe you don't talk to everybody every day anymore.
9:53
One thing that's present on our Kanban Board, it's really, it's probably hard to see on the screen is there's a little time stamp here about when did that thing end up in that column.
10:07
And so we could have like a date.
10:09
It could also say like a few minutes ago or hours ago or whatever.
10:13
So you can also get an idea of like how fast are we moving along or how long is this thing been in that state.
10:22
And maybe based on, you know, what you're seeing here, you can reallocate resources, move things on hold or de prioritise or like send it back to the team to revisit, like why are we making this compound?
10:35
Do we really need this compound?
10:37
Sometimes the answer is we have to have this compound and you're willing to wait.
10:41
But a lot of times compounds aren't that critical or any specific compound isn't that critical.
10:46
And so you can move it to some other lane here.
10:52
So this again, this is nice for the sort of the at a glance project status.
10:58
As you're going through the project, it would be good to know where you're at any given time.
11:03
But if we're going to think about that DMTA cycle again and like, how are we going to actually go faster?
11:09
We might also want to start to think about process changes to how we do things.
11:15
And in that case, you don't just want to have this like one little value about how long is a compound been in this status.
11:22
You probably want to have a summary of everything and how things moved along.
11:27
So we have a new feature that we've added to our tool as of the last few months.
11:33
And it's one of these buttons at the top, which will export as CSV file.
11:40
And that CSV file just has the raw data on compound progress.
11:45
OK, So what I'm showing here, this is a this is an actual, this is actual data from one of our partners.
11:54
It's all obscured.
11:55
There's no structures.
11:56
So but you can get an idea of like what this actually looks like.
12:00
So every row here is when material moved from one state to another.
12:06
And along with it, our IDs.
12:09
You can export smiles if you want to have structural information, the hypothesis and design sets that are associated with and links back to the entries and design hub.
12:22
And once we have this, you can start to think about how we want to process this and visualise this.
12:29
You could use Excel or spreadsheet, other spreadsheet tools or maybe 9 or Python, depending on how complicated or what you're looking for.
12:39
And this is kind of where we could maybe use the help from the community too.
12:44
We have some ideas about what we think interesting or useful reports and visualisation might be, but if there's things that you are doing and you like data that you would like to know about how your projects are going, perhaps this will be an entry into determining how to do that.
13:03
So on the next few slides, I'm just going to show some a few different visualisations of how we process this data.
13:09
This is probably this the simplest one.
13:12
All we've done here is we've looked at every compound that was in our data set.
13:25
And I've just ranked these things on sort of the slow movers and the fast movers.
13:29
OK, so in this case there's a compound that like almost took a year.
13:33
It was in synthesis for almost a year, some, you know, half a year and all the way down.
13:38
And some compounds even were in synthesis for zero days.
13:42
Maybe they're commercially available or you know, maybe we had them on hand already.
13:47
But some things of course can be made really quickly.
13:52
So you can start to identify maybe what are the kind of kinds of compounds.
13:57
If you drill down and you revisit like what are these structures that are moving really slowly, maybe we can figure out why some compounds are not going very fast and decide if we're making decisions appropriately, like is it worth waiting 324 days to get this compound?
14:12
Or maybe we should have cut bait earlier to focus on something else.
14:19
Another visualisation is what we are describing as the average compound progression.
14:26
So this is looking at how many days are compound spending in each phase of the process.
14:33
So if we're kind of maybe moving down the line, we have compounds that are like ready for review.
14:39
In this case there was some modelling associated with compounds.
14:43
So there's time associated with that.
14:45
Compounds of course, are going to be like ready for synthesis and then in synthesis and then things are complete.
14:50
We don't probably need to keep track of how long they're completed for, but we get that data too.
14:58
And you can see here like, OK, compounds were ready for review for like 0 days.
15:03
So we're making decisions at that point very quickly.
15:06
Modelling was requested quickly.
15:10
Modelling completed.
15:12
There was a quite a range of time that it took to complete that modelling.
15:17
Generally, things were not just sitting around ready for synthesis for too long, but of course, like the actual synthesis time took a reasonable amount of time.
15:26
And you can see a distribution on like how long are what's like what is the average or what is the 90th percentile of our compound sets and what's the low in the minimum.
15:42
Another visualisation is kind of getting a little bit more information about how long are compounds taking, like what is the complexity of our compound sets and how are we reducing that to practise.
15:56
So what we've done here is we've binned the all the compounds for the time that those compounds went from the review process to when we had them, when we had them in our hands.
16:13
So like many compounds or it's sort of like topping out at something like 30 days.
16:18
So you know, most cases the cycle time is something like 30 days.
16:23
But we have this like long tail there is this like one spike at 100 days.
16:29
Maybe there's some like systematic reason that we had some, we had a number of compounds that took 100 days.
16:36
Maybe there was, you know, Christmas holiday in there.
16:39
So things got slowed down but it all appeared at once.
16:46
One last visualisation that could be useful for like thinking of a project on a long time over like a timeline.
17:00
This is basically looking at sort of a [unclear] chart of how did compounds progress in relation to each other.
17:10
This is maybe hard to put all, you know, the full compounds set onto a slide, but you can kind of get the idea here that OK, some compounds started here.
17:20
Eventually they were completed.
17:22
Maybe there was other compounds that started around the same time and lasted longer in that before synthesis or in synthesis state.
17:30
And you can start to maybe even piece together like, OK, when we finished this compound, got the results back and that inspired the next compound.
17:39
And you can start to put together those insights.
17:43
So these are just a handful of the visualisations and ways that we've been working with these data sets kind of what's to do next or what are people doing or what are the insights that we can gain from these?
18:00
So we've been hearing from our partners that they're looking to drill down to what compounds and compound types and compound series are the most challenging.
18:15
Which of those are the bottlenecks for us.
18:22
One thing that I always wondered about is like, who are the best performing chemists?
18:29
Or is there are there any chemists that are maybe doing things in a way that are slowing down the process?
18:37
And we can also look for other bottlenecks and frictions specifically in the make phase of the DMTA cycle.
18:46
So I think that this data is going to enable project leaders to make those process changes to improve DMTA cycle times.
18:55
Again, the make phase often is the slowest step in that process.
18:59
So if we can speed that up, we might make a dent into the cycle times.
19:05
Currently Design Hub exports that data as CSV for offline processing.
19:11
We have some pre built templates in either Excel or Google Sheets.
19:17
If like you're interested in and doing some of this, doing some of this analysis, we can get you started with it.
19:24
But again, we're seeking community feedback and like what the most useful interpretations and visualisations might be.
19:30
And you know, if we find that they have some global utility, they could get added to Design Hub as native functions.
19:42
So I'll wrap it up there.
19:43
Again, thank you for hanging out with me.
19:46
As we're getting ready for lunch.
19:49
Please do stop by our booth at the exhibition hall on Andras Stracz, who's the principal product manager for Design Hub is here from Budapest.
20:00
I'm sure he'd love to talk to you.
20:01
I'm there.
20:02
A few of our account managers are there as well.
20:05
So thank you very much.
