Making multidimensional experiments a reality doesn’t just sit with one discipline.
Most commonly, it involves biologists and statisticians. So, what’s the key to open communication, and working together to do radically better biology?
That’s what we explore, in this season’s bonus episode, filmed at Synthace’s Beer, Bytes and Biology annual event.
Conversation highlights
00:00 - Introduction
00:49 - Condescending attitudes causing statisticians and biologists to butt heads
02:49 - The key to successful relationships between the 2 disciplines
04:09 - Markus’ experiences showing how he and his team bridged the communication gap
08:39 - Embodying the “George Box” attitude for successful collaboration
If data is the new oil in an AI centric world, then amassing ever-larger multidimensional datasets can only be a good thing.
But how can we use these datasets to gain deeper insight into biology?
This is the question we try to answer in season 1’s final episode. Between us, we bring views of the AI optimist and skeptic—with starry-eyed visions and sobering realities—to the table before reaching a conclusion.
Conversation highlights
00:00 - Introduction
01:18 - A vision for multidimensional datasets fueling better experiments
03:37 - AI and myth-busting: “It’s not magic”
09:15 - Risks of combining datasets for making “pizza” and “ice cream”
10:11 - Ideals and potential low-hanging fruit for AI in biology
13:09 - Thinking about AI as collective memory
18:16 - Combining human and artificial intelligence is key
To explore what the future of multidimensional experiments might look like, we decided to look back.
In this episode, we explored how different multi-dimensional (aka Design of Experiments, or DOE) methods have come about to date.
Then, we pondered how these different methods, together with tech and software innovations, have brought both opportunities to scale multidimensional experimentation, and exciting questions on the best way to go about it.
Conversation highlights
00:00 - Introduction
01:04 - The question mark around automation: what to do with 1000+ runs
10:15 - History of multidimensional (Design of Experiments / DOE) methods to date:
10:58 - Era 1: Factorial designs came about, and benefited agriculture
15:12 - Era 2: Response surface methods emerged in process industries
18:17 - Era 3: Software packages helped design the “optimal” experiment
21:29 - “DOE 4.0”: Bayesian optimization, and the pros and cons for biology
In this episode, we delve into Markus’ experiences with doing multidimensional biological experiments manually—from the exhilarating progress he made, to the definitive results he produced.
Plus, we touch on how automation can scale multidimensional experimentation, and when is the right time to bring it into the mix.
Conversation highlights
00:00 - Introduction
01:24 - Markus’ early manual, multidimensional experiments, and their definitive results
06:07 - The overwhelming number of combinations you explore in biology vs chemistry
11:39 - What happens when you don’t use multidimensional methods
20:37 - When you should automate multidimensional experiments
23:02 - The exciting, uncharted territory that automation brings
It’s easy to say that the tried-and-tested way of doing biology isn’t helping us progress.
It’s quite another to embrace new approaches.
That’s what we’re covering in this second episode. We talk about communicating the power of multidimensional experimentation for biology, the insights they unlock—and how often, it takes some time to entertain new-and-improved ways of working.
Conversation highlights 00:00 Introduction
01:36 Phil’s tale of Dr Stevie vs Dr Charlie, or traditional vs multidimensional methods
10:40 Multidimensional? Design of Experiments? DOE? It’s all one in the same
12:19 How Markus got to a 7-fold increase in 3 weeks using multidimensional methods
15:55 The magic of multidimensional experiments lies in the statistics
16:49 Markus’ envelope-pushing multidimensional experiment with 27 factors
20:00 Markus admits that at first, he dismissed multidimensional experiments
In our first episode of The Next Experiment, we start by unpacking that all-important question:
Why is biology so hard?
In order to answer it, we get into the fundamentals. The nature of nature.
We talk about how biology’s interconnectedness makes experimentation in biology so uncertain; why the standard method of varying one parameter at a time isn’t cutting it; and how switching to a multidimensional approach is more important for biology than any other scientific discipline.
Conversation highlights
00:00 Introduction
00:45 Why biology is different from other scientific disciplines
03:40 What emergence means, and how it relates to biological systems
06:03 Chemistry is a “solved problem”, and biological systems are “black boxes”
12:18 How multidimensional methods helped Markus make definitive progress
16:50 The counter-intuitive science lessons Phil remembers from primary school
We’re Markus and Phil, a biologist and statistician.
We decided to come together and discuss the best experiments to cut through biological complexity.
If you have this sense that there might be a better way of making progress in biology, subscribe and join us.
Stay tuned for our very first episode, launching on November 6th.