0:02
Thank you and thanks to Brooks for being a sponsor and inviting me to speak.
0:07
It was really, I think Mike Warren that thought of the idea for me to speak here.
0:12
So thanks.
0:13
Mike and I worked together on various projects for I don't know if Mike wants me to say how long, but I think 25 years or so at different companies.
0:22
So he kind of knows what I'm about what and what I've done and asked me when we talked about a topic, you know, I said, well, I could give my company pitch deck or something like that.
0:32
And no, I don't really, I don't think people would really appreciate that.
0:35
And then we kind of came up with the idea of just a slice through my career of the type of automation I've been exposed to how it was implemented and the evolution of that.
0:45
And I've given a lot of talks over the years.
0:48
And a lot of my slides behind each slide would normally be about 30 slides.
0:52
And so these are 30 to 50 minute talks, and I have 25 minutes, so it's going to be kind of high level.
0:59
I want to leave some time at the end for a question and answers.
1:04
So agenda.
1:05
Well, first, you know, I'm going to talk about my experience, early stage drug discovery and sailing.
1:12
And why am I going to talk about sailing and why am I wearing a Hawaiian shirt That will come to be known.
1:18
I think a couple people in the audience know why, but I'll get to that.
1:22
That'll be very obvious.
1:25
I'm going to talk about early automation in my career, what that was like, how workstations came into being and also dedicated equipment, very high throughput equipment.
1:34
And then versatility was brought in.
1:35
And then there was a move to industrial scale and very high throughput.
1:39
And I'll talk about that a little bit.
1:40
And then in came collaborative robots in the lab, working with people.
1:45
I'm going to talk about the impact of mobile robots a little bit.
1:48
I mentioned that earlier.
1:49
And also finally in, with miniaturisation of multiplexing in the impacts of that.
1:54
So in the, in at the very beginning of my career, about this time, early 90s, I graduated undergrad and I got a job as an internship at Genentech.
2:04
And that was just great in the hybridoma group at Genentech.
2:07
So it was lots of cell culture and things like that.
2:11
They ended up picking me up as a research associate and I worked there for three years.
2:15
But the automation in the lab was, you know, it's a pipette man, it's multi channel pipetter.
2:19
We had a couple thing instruments like a spectrophotometer and then eventually we went to a plate based special absorbance reader, you know, things like that.
2:27
We actually had a stacker on our absorbance reader.
2:29
There was AI got a lot of exposure to things like flow cytometry, bioreactors, Western blot apparatus, things like that.
2:38
It was those type of things.
2:39
Then I moved to San Diego Ligand Pharmaceutical, still in the early 90s, early and mid-90s doing more molecular biology type of techniques, but still the equipment was quite similar.
2:49
Not anything real high throughput, but while I was doing this, I kind of got obsessed with the idea of sailing a boat single handed from San Diego to Hawaii and Hawaii to San Francisco and back down here.
3:04
And I became very obsessed with that idea.
3:06
Got an old boat, fixed it up and took a leave of absence.
3:12
So I'd worked a few years.
3:13
So I took a leave of absence to really understand what it takes to sail a boat.
3:17
And so I sailed around Southern California for a few months.
3:21
This was an old boat that I got and fixed up.
3:23
So San Diego's down here.
3:25
You see Point conception up there.
3:26
That's about 200 miles.
3:28
So I explored pretty much all these islands.
3:31
San Clemente's military, you can't go on that.
3:33
Avalon and Catalina, I think people know about that.
3:35
And then there's a little speck of an island called Santa Barbara out there.
3:39
And then all the upper Channel Island, so Anna Kapa, Santa Cruz, San Miguel, Santa way out at Santa Rosa, there's a Kyler harbour is a really nice spot, and you see the boat there.
3:48
I fixed up.
3:50
So anyway, went back to ligand and I was on a leave of absence, but they didn't keep my old position.
3:58
They said we have another position for you and it's in with the department.
4:02
They called new leads.
4:03
So new leads did high throughput screening.
4:05
What was high throughput screening at the time?
4:07
And also Med chem SAR support and the people who had basically brought that lab to be had all left.
4:13
And so there were operators, users, but no one really understood the technical challenges.
4:17
And the director said, well, you fixed up your old boat, maybe you can help us here.
4:22
So I went in there, they had all these workstations.
4:24
These are the actual, we're pretty closely the actual workstations we had most of them just did kind of one thing.
4:29
I started going to lab automation conferences and stuff like that, and it got exposed to the concept of the laboratory unit operation.
4:36
So that's a dispense or a wash or a read or something like that.
4:40
And you link your LUOs to make a method or a process.
4:44
Exception being the Beckman Biomek 1000, which was more versatile but at the expense of throughput.
4:54
At the same time there was a thought to get higher through but so a company called Zymark out of Boston came up with the Allegro system.
5:03
And Zymark was very progressive automation back in the 90s and they would make, you know, they had robots and kind of semi anthropomorphic robots, but they came up with this Zymark Allegro system to try to break screening 100,000 compounds a day.
5:18
That was ultra high throughput.
5:19
Consider ultra high throughput.
5:21
And if you'll notice that it has several boxes inside each box is 1 instrument.
5:26
And so when you develop your process, you would for each step of your process, put in a box and then there's an instrument in there.
5:34
Say for instance, you might have a dispense box, an incubate box, then another dispense box, a plate wash box and a read box or something like that.
5:42
You can incorporate compound transfer in there.
5:45
So for your, you would run your high throughput screen for a couple or a few weeks and then if you got a new screen or a new process, you rearrange the modules.
5:54
So those modules are on wheels and you'd actually rearrange them and then validate your process and go again later on.
6:01
They did get a little bit more versatility where they could pass things back and forth, but they did hit 100,000 a day in 384 well plates they were able to do that.
6:13
A slightly different approach is this system that I actually became responsible for back at Ligand.
6:19
It was on the cover of the journal Laboratory Automation in 1996.
6:24
The folks that really built it had left the company.
6:27
So this was left to me to figure out.
6:31
But it has a robot on a track.
6:33
It's one of the very first robot on a track systems and you see all the different instruments around the edge.
6:39
And so it actually was quite versatile because and you could programme it, you didn't have to rearrange modules.
6:43
You could actually programme this to operate in a different order.
6:48
And I got to know every instrument on there really well because this is at back in this time, these instruments generally weren't designed to run this hard for, you know, to run outside normal business hours and so on.
7:00
And the software or the company CRS Robotics in Canada really embraced helping me get the system back up to performance.
7:10
And they had me go out and give some talks on it and things like that.
7:13
And it was a very nice experience.
7:16
But in the meantime, I really wanted to go on that sailing trip.
7:21
So this time I asked for a leave of absence and they said no.
7:26
So I quit and I had this great sailing experience.
7:30
So I sailed from San Diego to Hawaii.
7:33
It took three weeks.
7:34
I spent three months sailing through the Hawaiian Islands.
7:37
So from Hilo on the Big Island to the other side, Kona to Maui, Molokai, Oahu, and finally Kauai, where guys from we've worked together for a long time and then sailed from Hanalei Bay back to San Francisco.
7:54
And you might wonder why it's such a big route around like that.
7:57
That's because there's a weather phenomenon called the Pacific High.
8:01
In the Pacific High in the summertime is up north, the winds go clockwise around a high and so it's a perfect sail to Hawaii.
8:08
So the races all happened in mid Midsummer.
8:10
The sailboat races. I went in the winter because that's the time that was a time I had to do it.
8:16
So you can see my route went S because the high was went down South.
8:19
And that's why we get the weather in the winter.
8:20
The storms come over the top of the high, but so I had to go pretty far South.
8:25
So it really extended the trip.
8:26
And then coming back North you go around the high.
8:29
I didn't say in the middle of the high has no wind.
8:32
So you want to avoid the middle of the high.
8:34
And so you see myself sailing straight north and I went into the middle of the high right there and I motored across the high for about I think 24 hours.
8:43
And then I got to the other side of the high and the winds went the other direction and I sailed to San Francisco.
8:48
What's interesting is when you're in the middle of the Pacific like that, it's equal distance to Ketchikan, Alaska as it is to San Francisco.
8:55
And I really wanted to go to Ketchikan, but the boat wasn't up for it.
8:59
So I kind of attenuated that a bit.
9:02
So I still have a little bit more to go.
9:04
But I was very happy.
9:06
I achieved my great goal, my obsession, and I went had to go back to work.
9:12
So just this is 99.
9:16
I went to work at Novartis.
9:17
This is the year the institute was founded.
9:20
I think Guy and I started their same time, right?
9:24
You were there before me a little bit, yeah.
9:25
So but the whole world had changed.
9:29
When I came back from that ship, it seemed like the world had changed.
9:31
Everything was high throughput.
9:33
The genome was about to be sequenced.
9:35
We thought we were going to get, you know, hundreds of drug targets if not more out of that and lots of pharma, some biotechs were setting up to be able to screen hundreds of targets.
9:44
And certainly that was the view of Pete Schultz, the founder of this institute.
9:48
So he hired these industrial type engineers to bring in industrial approaches, and we end up working with these stably robots that we got from Mike Warren back there and put that together, this ultra-high throughput screening system.
10:02
We've revisited everything on the system to make it very high throughput, broke the workflow up to three robots.
10:08
We meant also miniaturised.
10:10
So we started working in 1536 well plates instead of 384 well plates.
10:14
We had to develop some technologies to keep the cells happy in those low volumes, but that worked.
10:21
Another concept was storing the entire compound library on the system.
10:24
So when you went to screen, you didn't have to move compounds around on and off the system.
10:29
You just loaded assay reagents and plates, and you would execute your screens.
10:32
And we went well into UHDS territory, which is 100,000.
10:36
These could actually run several 100,000 wells a day if that's what we wanted to do.
10:41
We spun out a company called Calypsis.
10:44
So Calypsis took advantage of this, these robotics for their own drug discovery, but they also commercialised them.
10:51
And I got involved with the commercialization of these systems.
10:53
So I got to go all over the world pitching these.
10:56
And generally they went to things like research institutes and very large pharma companies.
11:02
NIH got one.
11:04
It's public.
11:04
I think they paid $27 million for it.
11:08
Merck bought three of them.
11:10
Sanofi got one.
11:11
BMS got one what a government of Taiwan and Novartis got a few scattered around.
11:20
They never really got traction in Europe for some reason.
11:24
Asia, United States, they did.
11:25
And we did sell some of the dispensers to places in Europe such as Weisman Institute.
11:32
So I got to go to Israel and go to Jerusalem and things like that.
11:37
So it was a very nice experience for me.
11:42
How am I doing on time, by the way?
11:47
Good.
11:48
OK, Ten.
11:50
OK, Eight.
11:51
OK Good.
11:51
Great.
11:52
OK, So what next?
11:55
Well, I did get married.
11:59
We incorporated the sailboat into our wedding reception and then took a short holiday and then had to get back to Novartis to work.
12:09
About that time something came on the scene called collaborative robots.
12:15
And collaborative robots are robots that work in the same area as humans.
12:20
They're basically not dangerous.
12:23
The industrial robots needed extensive gardening.
12:27
They took up a lot of space.
12:28
Very reliable, very fast, but took up a lot of space.
12:30
These collaborative robots can go into laboratories.
12:34
They can work with people in the lab.
12:37
You can use them various ways.
12:39
One, and this is actually in at a Stratios lab I was responsible for in Menlo Park.
12:48
They have facilities in Menlo Park and San Diego.
12:53
It improves space utilisation relative to having instruments spread around your lab and humans going to them.
12:59
You can tighten up the footprint and have the instruments in nice and close.
13:05
You can also use collaborative robots on call it islands of automation or modules where you have a number of instruments.
13:12
Here's a, there's a track on this one and you can execute screens kind of like the robot I had at Ligand on a track.
13:19
But these are collaborative, so you can have them be closer to people.
13:25
Now, Stradios had two locations that I was responsible for, one in Menlo Park, it's this one.
13:29
And then the one in San Diego that was actually owned by the instruments were owned by Eli Lilly and Stradios operated them, developed the software and operated them.
13:37
It had 500 feet of track and automated chemistry, purification, sample management, analytics and screening of various types.
13:49
It had about 20 modules on it connect all connected by 500 feet of track and puck.
13:55
So the process would move from synthesis to purification to sample management and then out to the screens.
14:01
And that worked.
14:02
And the software is very sophisticated, and we were working with that.
14:06
The interesting thing was that kind of came up was when it's 500 feet of track, it's about 200 feet end to end.
14:11
If you're in the middle of it and you have to go 5 feet on the other side to work on something, you have to walk a couple 100 feet around.
14:18
So you know, looking back, that might have been a good opportunity for mobile robots.
14:24
So mobile robots are kind of coming on scene now.
14:29
I've never had a mobile robot, and I've been kind of looking for that killer application.
14:36
So I mentioned that one that possibly could be, and that's why I asked the previous speaker, have you, are you thinking about mobile robots?
14:46
Yeah, some of the considerations about mobile robots, you know, they're considered collaborative.
14:50
They have sensors on them.
14:52
So if they come to your person, they'll slow down.
14:54
But if they're in a lab that has a lot of people, robots have right away.
14:58
So they're going to be stopping all the time.
15:00
And so the throughput probably is going to be very good.
15:02
So that I don't think will be the use of them, but I think connecting islands of automation could be.
15:07
So you have certain operations occurring on modules and then you connect them because occasionally they have to move plates back and forth.
15:21
Next is where I am right now, Plexium.
15:25
So at Plexium we have a proprietary technology.
15:30
Peggy Thompson's going to be talking about it more this afternoon I suspect.
15:34
It starts with a bead, the bead has on it.
15:36
It's a 40 Micron bead compound with a photo cleavable linker.
15:41
It also has a DNA tag that indicates what compound is on that bead.
15:46
We put those beads into our proprietary microfluidic devices, which I have one right here.
15:52
This credit card size device has 88,000 wells in it.
15:56
We run cell based phenotypic assays in this.
15:59
So we put the beads onto the microfluidic device.
16:03
We have a technology so we with very high efficiency get one bead per well.
16:07
We then put in the cells, let them settle down a little bit, photo cleave the compound, so it now separates from the bead and the DNA where you know, traditional DNA encoded libraries, the DNA barcode and the compound stay linked here.
16:25
The compound separates so it's free to cross the cell membrane going to cell and have an effect.
16:29
We're at the degrader company.
16:30
So we tend to look at date at depletion of proteins.
16:36
You can measure those with immunofluorescence, you know label fluorescently labelled antibodies.
16:41
We then determine the positives with our proprietary high content imager.
16:46
It's an imager that has four flush channels, Brightfield and Darkfield.
16:51
It can read this plate 88,000 wells and six channels in about 10 minutes.
16:56
With a regular image high content imager that would take several days of 3/4 well plates.
17:04
So we determine which wells have the positives, we then pick them and then decode them with next Gen sequencing where we can do offbeat macro synthesis to confirm the activity.
17:15
So some of the things we measure for quality during this process, we always monitor the beads and picking efficiency and of course assay Z squares.
17:24
For morphology metrics.
17:25
We of course follow intensity like I mentioned, but it's a high content imager, so we can perform cell segmentation and then all the usual high content imaging metrics.
17:34
As far as cell types, we really haven't come up against any cell types we haven't been able to use.
17:39
We've done everything from cell lines to an advanced neuronal assays.
17:46
The reason we can do that is because we have proprietary coatings in the devices and the and we can change and modify those coatings for particular cell lines.
18:02
To increase reliability to the system we came up with this fluidic module so we can control all the micro fluidic reagents very precisely as opposed to pipetting them manually.
18:13
So we can wash the cells, fix the cells, stain the cells and so on.
18:21
We established 1 footprint for our microfluidic device.
18:26
We actually have a couple of different variations on that.
18:29
So for going fast and working with cell lines, we have what we call PX 26 that has 88,000 wells.
18:35
For more advanced or complex assays, we have the RPX 27 which has fewer wells, but they are larger volume.
18:41
And so some cells are happier in the larger volume.
18:46
And I talked about our proprietary high content imager.
18:50
Oh, so some metrics, they're just kind of for fun.
18:53
188 thousand well device to get that many wells in 384 well plates would be about 230 plates.
19:00
It takes 1.5 mills to fill our device with reagent.
19:05
It would take several litres of reagents to dispense into a 384 well plates.
19:11
So we're decreasing the cost, increasing the throughput.
19:14
But what's really enabling is we can run screens at a scale with rare cell types or very expensive reagents not otherwise possible.
19:26
And the time to run a 500,000 member screen, it's about two weeks or 4 weeks, 2 weeks for the simpler assays and four weeks for the more complex assays.
19:38
So what's next?
19:40
Well, we've been hearing these last couple days a lot about AI and how AI is going to help with drug discovery.
19:48
Just looking at automation, you know, thinking where AI is going to impact there.
19:51
It could, you know, on the design and deployment of the systems for optimising the workflows.
19:57
I think AI could have some impact there and to automatically manage and recover from errors and to basically, you know, get feedback from these robotic systems and continuously and improve the process.
20:12
Also more advanced applications, you know what I see really coming is full spatial proteomics, transcriptomics in more biologically relevant 2D in 3D biostructures.
20:23
So that is, you know, could be organoids.
20:28
But really looking at that, you know, we have the ability now there have been several papers that have come out to get full proteomic, well not full transcriptomic and some proteomic signatures.
20:39
And then multiomics is when we bring those together, we get proteomics, transcriptomics, maybe metabolomics all at the same time.
20:49
And the last thing is automated high throughput chemistry.
20:52
So a lot of what I just discussed is get you some hits and then you confirm the activity of those hits.
20:59
You can train your AI model, whatever, but then you need to synthesise compounds and test them again.
21:04
And that's pretty much always a bottleneck.
21:07
I saw yesterday E molecules had a nice presentation where they talked about their latest automated chemistry that they're implementing.
21:17
And they're a particularly good place to take advantage of that.
21:21
I've worked with the in the automated high throughput chemistry before and often it's waiting for those intermediates that is actually the limiting factor, not necessarily performing the chemistry itself.
21:30
And since they're E molecules, they have the intermediates.
21:37
So that's it.
21:38
That was my career in 20 minutes.