0:00
Thank you for coming today.
0:02
If you can't hear me, then you're deaf because I'm exceptionally loud.
0:07
I'm going to be talking today about FAIR.
0:10
Now maybe if you came to this talk, it's because you love the word FAIR or you actually are one of the 28% population of researchers that actually know what FAIR is.
0:24
We got some nods.
0:25
OK, there's some FAIR fans here.
0:26
That's great.
0:27
If you're not a FAIR fan, we're hoping you will be by the end of this talk and then the end of our panel discussion directly after this.
0:36
So as already mentioned, I'm a former researcher.
0:42
If I had FAIR at the beginning of my career, my data had been structured in a meaningful way and we hadn't just started adding tools to our tool belt, then my life probably would have turned out very differently.
0:58
But I got exceptionally frustrated about how that was managed and ended up starting a software company to solve that problem.
1:09
What I realised very quickly is that when I learned about FAIR, which was well after I started my software company, that we had gone digital.
1:20
And it was absolutely incredible how digital we became.
1:24
But digital does not mean FAIR.
1:26
We added ELNs, LIMS, quality management systems, but as a scientist, I'd look at my data from a year before and yeah, I'm still trying to figure out exactly what context was, what the experimental conditions were, all of the factors page by page, flipping through notebooks, going back to Excel sheets.
1:51
It was exceptionally difficult.
1:52
Also, compliance, anything we wanted to do in the US to try to get a drug towards FDA and our partnerships was exceptionally difficult.
2:02
And that wasn't because we didn't have technology.
2:05
We had lots of technology.
2:06
I mean, we were in the rush of technology at the time.
2:09
It's because we didn't actually intentionally design what we were doing.
2:15
We designed for the specific problem and then stacked those together.
2:22
So let's go back to what it was like.
2:25
Scientists re-entering data across systems.
2:29
Excuse me, I'm going to clear my throat.
2:39
Thank you.
2:41
We have EHS and lab ops people.
2:44
It's a very integral part of good management and good operations is a lot of the data comes in and goes out for different purposes and lab ops and EHS are key component of that.
2:56
If you don't have that integrated into your programmes, then somebody, and most likely us as scientists are doing double or triple that entry or managing separate spreadsheets, probably not for only us but for other people.
3:12
And then you've got our IT leads.
3:13
We bring in all these IT leads, which is absolutely great, but what they'd end up doing is taking these older systems and trying to patch them together with APIs, they're trying to create additional workflows because they can't do something.
3:30
So they create another workflow.
3:31
And now as a scientist or as somebody having to deal with the data you're managing and trying to integrate all these different components, being that IT person, that is an incredibly stressful job.
3:45
So I would like to step back from things and look at it from maybe a different perspective.
3:51
We hired all these amazingly talented musicians and we got them those fine instruments, the, you know, the beautiful quality French horn and the violins.
4:04
And then we didn't bother to bring in a conductor.
4:07
So we should not be shocked that what we were hearing a lot of is noise and not the beautiful Symphony music that it could be.
4:17
So the dilemma is how to tackle this, right?
4:22
A lot of organisations are picking one of two strategies right now.
4:27
And you may be from one of these organisations that's doing this.
4:31
On the first hand, you have aggressive consolidation.
4:34
What does that mean?
4:35
That means telling everybody, we know you probably have your tools as you like, but this is not feeding our ability to do new, maybe AI Gen focused research.
4:51
So we're going to consolidate you all into one tool or one, you know, small set of tools.
4:58
So that's on one.
4:58
The other is freedom for all, right, which I loved.
5:03
I loved as a scientist, I had enough funding that if I wanted a piece of software, I could just pile it back on top of my stack, you know, just add one more thing, let somebody else's problem.
5:14
And of course I wasn't going to do anything with the data after I published, right.
5:18
So wasn't a big challenge.
5:19
But that is that those are really the two systems that we see right now.
5:28
So now of course, we have FAIR.
5:32
Now looks like we've got a fan of FAIR here.
5:35
But if you don't know much about FAIR, just make sure we go through a few things here.
5:40
So findable, accessible, interoperable and reusable.
5:46
This was based on a paper that was published in 2016, again after I started my company.
5:51
It would have been great had I known about before.
5:54
It was based on a workshop that was done in 2014, but still in 2025 or I guess 2024, the last survey that I saw about it, only 28% of scientists knew about it.
6:06
Doesn't mean they weren't practising some parts of it.
6:09
And that's really important to understand.
6:11
Fair is not an all or none mentality, but it does mean that it's not embedded in our culture.
6:18
So it's something you build, it's something you architect.
6:24
And let's talk about some of the reasons why I think most people are here because they understand these reasons and we'll go through a little, a few of them actually, there's a lot of them.
6:37
So first and foremost, reproducibility is the biggest issue still in research.
6:43
I'm sure there's some others, but reproducibility is way up at the top.
6:48
So redundant research.
6:50
How many pharmaceutical companies take a take an early study and have to actually repeat the entire thing in house because whoever they partnered with originally didn't have the data structured in a way that they could really question it?
7:09
It happens all the time.
7:10
It's something over 30% of all first trials are repeated for that exact reason.
7:19
OK, we have drug discovery.
7:21
We have collaborations and partnerships.
7:23
It doesn't matter if you're from academia and you're trying to partner with the commercial side or you're from the commercial side and trying to partner with academia.
7:31
If you don't have a great data structure, you're not going to be successful at that.
7:35
It is a major, major problem.
7:37
And if you want to invest in that, then investing in FAIR, be it full blown or even kind of a light version is going to improve your ability to collaborate or maybe outsource.
7:50
If you're a smaller group and you don't want to do all your research and you're going to hire CRO, you're going to need to be able to structure data in a way where that CRO can actually get something out of it for you and doesn't spend half their time either redoing the research or going back and cleaning the data up so that they can do something with it.
8:09
AI, everyone's talking AI.
8:12
Well, if you had million dollar budget for AI, most people would be using 50% of that right now to clean up their data so that they could use AI to actually ask questions.
8:22
That is not a wise investment.
8:25
So what you really want to do is invest in having your data structured so that you can get the most out of your AI budget.
8:34
Regulatory readiness.
8:35
I think we already hit this one.
8:37
Rare diseases.
8:38
Rare disease research rarely happens in one organisation.
8:41
It's usually a collaboration of many small groups across the world.
8:46
And you need to be able, you don't have enough data cause rare diseases, you know, produce a fair amount of data.
8:54
So what you really need is to be able to pool those data in a meaningful way.
8:59
And you cannot do that unless you have a proper data structure to work from.
9:03
And then, you know, as we continue to do research, especially larger organisations, you build up institutional knowledge and institutional information.
9:13
And if you can't mine it, you can't utilise it.
9:16
You're actually just paying for a whole bunch of data storage.
9:18
So you might as well structure it in a way where you can get something out of it in the future.
9:23
And what that is obviously going to depend on your research.
9:28
This is the negative part of the presentation, OK.
9:32
There are a lot of reasons why verification fail and the whole goal here is not to make you feel depressed because this has happened to you, but to help you look out for what could happen if you're trying to do this at your organisation.
9:49
So let's get into it.
9:50
There's three main categories.
9:52
There's technical myopia, there's misaligned design and there's organisational neglect.
9:59
Does anybody have any favourite ones here that they've experienced at their organisation?
10:07
Somebody's got to have at least one.
10:09
There's 10, so maybe it wasn't data-driven.
10:14
You got something.
10:20
So one I'll highlight is ignoring culture and incentives.
10:26
So somebody comes in, says we're going taking on a new tool, a new structure, and they push it down and what do they get?
10:35
They get a whole bunch of chemists saying this was designed for biologists, a whole bunch of biologists saying this was designed for chemists.
10:42
A lot of water cooler talk.
11:02
So if you didn't hear that, what they mentioned was technical myopia and just the over reliance on the technology, the over reliance on the tools and the integrations and the components and not on enough of the other components of what it takes to implement verification in your organisation.
11:23
OK, so we basically hit this, but it's cultural and architectural.
11:30
It's not technical.
11:31
We got plenty of technical.
11:34
So what do we do?
11:36
What could we do differently?
11:38
OK, I don't have the only solution.
11:40
I have a view on it.
11:43
So I believe in shared metadata layers modular systems with open APIs now this open APIs not open science necessarily, right?
11:55
The view is though, that you should be able to, with whatever systems you're working with, you should be able to use your data across and that should enable you to tie in a shared data layer.
12:08
Domain specific.
12:10
I'm a true believer that you cannot put all scientists in the exact same interface and expect them to thrive.
12:18
Scientists needs flexibility.
12:20
They absolutely need flexibility.
12:22
So domain specific, researcher first interfaces, the interface matters.
12:27
It really matters a lot.
12:29
And if you want to culturally engage your community, you have to engage your scientists.
12:35
Because I remember when I didn't like something, I just didn't use it, and they couldn't do anything about it.
12:42
So maybe that's not the mentality anymore, but I don't know.
12:47
I think there's some people here who agree with that. Compliance is critical.
12:52
That should be part of your design governance.
12:56
I think my most the most important one is this infinite loop implementation.
13:00
A lot of people do a one and done implementation.
13:03
They bring in A-Team, maybe they have a group assigned internally and they do an implementation, and they take three months, and they do it.
13:11
And then one year later, 20% of your researchers are different.
13:19
And that 20% was never trained, was never bought into a culturally, there were never incentives for them.
13:26
And so if you're going to go about this, I recommend having a continuous loop where the implementation people come back in and pick it up, where you provide guides and videos that make sure you're it's part of your onboarding process.
13:43
But don't just count on HR doing it.
13:46
You actually have to have implementation people come back in and rerun a mini version of it.
13:51
So, OK, so how does this actually play out?
13:57
So here's this, here's a use case, OK, here's a situation of a large R&D organisation that had, I don't know, around 200 labs.
14:09
And they had a big problem, OK, about 1300 people, I think.
14:14
And they had a big problem.
14:16
They're trying to get more out of their data, but they ended up having, they had disparate types of research across this large organisation.
14:23
Not, you know, you got your certain types of organisations.
14:27
They kind of they're in their lane, they can share data.
14:30
The same metadata structure that somebody decided on early on still works because the research is more narrow.
14:37
But this is a much larger organisation that has very diversified research.
14:42
OK.
14:42
That's very true about your big pharmas and it's very true about your big academics, OK.
14:47
And they had a big reproducibility problem, which is not new, but sometimes this also gets you in trouble when you get audited.
14:55
So what they had to do was start pulling their act together.
15:00
So the goal was to align the data practises across all of these groups, which is really hard when the type of research is diverse.
15:13
The goal was to centralised the metadata you have, but I think the biggest challenge is when you try to do that, you normally only are willing to do that when you're ripping out existing systems and putting new ones in.
15:28
So this was something that this organisation wasn't going to fund.
15:32
They weren't going to fund ripping everything out.
15:35
New connected systems, you know, new equipment that automatically integrated because of the preferred partner relationship they had with their vendor.
15:45
They weren't going to do that.
15:46
OK.
15:47
So what they had to do is have a really simple system where they could integrate. What happened?
15:55
Well, they ended up spending a lot less time on data management.
16:02
The collaboration went up dramatically.
16:05
And I think that was one of the coolest things, labs could actually pull together their different research.
16:11
And we're hoping that turns into real long time, long term sustainability.
16:18
OK, so let's try to wrap up here.
16:22
I know we're right up against the end here.
16:25
What is adopting FAIR do?
16:27
It speeds research.
16:29
It reduces compliance risk.
16:32
Compliance may be an internal thing at your organisation and may be a big external thing that you have to tackle.
16:38
Maybe you're taking drugs to market, maybe you are filing, you know, for certain types of embryonic tissue usage and you have to keep track.
16:51
Well, people aren't just happy with the same old look at my lab notebook view anymore.
16:57
Valuation and partnerships is critical to have good IP management, and it turns your data into a real asset and not just a cost.
17:09
So tomorrow's scientific leaders will not just look at their tools and try to figure out how to make the best of it, they will design it.
17:21
They will sit down with their scientists.
17:23
They'll sit down with their data leads, and they'll find a way to architect it into their culture and into the way, I guess the culture is the way they work, the way they operate and have shared purpose around doing this.
17:38
Thank you very much for your time.