Bioplatforms Australia's Landscape Analysis of AI - generative AI's impact on biology research infrastructure
- Steve Quenette
- Mar 21
- 4 min read
Updated: Mar 26
The rapid developments and availability of Artificial Intelligence (AI) technology, alongside associated growth in computational resources and data, have stimulated the widespread adoption of machine learning-based methodologies across research, industry, and society. What is generative AI's impact on biology research infrastructure?Understanding the global AI-based technological change within the biosciences is critical to the Australian community.
As one of the NCRIS-funded National Research Infrastructures, Bioplatforms Australia sought to understand the impact of the #generativeAI era on the #omics / molecular life sciences community. They engaged us to consult with their stakeholders and ultimately provide preliminary advice and recommendations on a framework investment. Specifically, our task was to prepare BPA for future investment, test preliminary stakeholder ideas, and establish principles for prioritising high-impact initiatives. For example, #AlphaFold has captured attention and been integral to changing beliefs - we validate that and explore where the next AlphaFolds will come from and what role Australia plays in creating and using them.
We shared some preliminary findings in September last year. See the news piece: AI driving sentiment change: biology has changed from hypothesis-oriented to engineering-oriented. Earlier this year, we circulated a public report to the BPA stakeholders: BPA landscape analysis of AI & AI-era strategy - consultation & strategy summary.
This summary aims to disseminate some key findings and, in turn, empower the community to progress with well-considered AI-era agendas.
The report suggests a vision that BPA and its stakeholders could adopt for the AI era (the next decade). This vision builds on the sentiment change and ties in the key points: industry is out investing academia, and biology is a small data discipline:
The generative AI era will decipher the language of biology.
The globe is embarking on a grand challenge - to “decipher the language of biology”, not dissimilar to when the discipline embarked on discovering the human genome. It is paramount that Australian research and industry participate and yield sustained value from this.
Within the AI era, the research / industrial complex will claim to have solved the central dogma of biology. We'll know how to model and predict DNA, RNA, and protein interactions. With that, biology truly becomes a predictive discipline. AI enables us to #tokenize (make words out of) DNA, RNA, and proteins based on our ever-increasing and gigantic source of omics data and our ability to align that data with broader health/phenotype data. As the key partner in facilities generating omics data, BPA is ideally placed to drive ecosystem change. Researchers and innovators will then use these words to create sentences to convey a message - perform some specific biological function. The race is on to discover these languages and use them to accelerate the process of drug discovery, personalised medicine, and better understanding of disease. Moreover, this is not just a one-off technology purchase; it's about tying your rate of discovery/invention to, at worst, Moore's Law and, at best, the 1000x speedup improvements seen in AI.
If you believe in this imminent disruption, what are the capabilities, and who are the partners you need to get there?
When observing the landscape from this perspective, another facet becomes clear. The AI era unlocks investment from those who are not proficient in biology. For decades, investors have understood how to measure outcomes from an engineering process (process power / scale economies / etc.) and used that knowledge to invest in markets.
Another key consideration is that the research and industrial community hold small data. The outputs from omics instrumentation are small compared to global knowledge of the biological system. The key to success is safely integrating your small data with global knowledge. This is the clearly tricky part, but when one uses an LLM, this is essentially what one does — we continuously integrate small data with global knowledge. We must create a culture for continuous integration and hence create the factories to do so.
We engaged 231 researchers, research enablers (capability providers to researchers) and those formerly in the research sector.
110 directly engaged, of which 64 led to direct contributions
we held 3 events, incubated 14 ideas and engaged 5 big techs
Some of the challenges to wide-scale adoption are:
Understanding AI-era business models and opportunities that give rise to researcher-led ventures
The paradigm and skill to succeed in the AI era are different the traditional (medical) research ways
AI infrastructure is (presently) scarce, and the software is evolving quickly - access to pilots can be cheap but at the expense of margin at scale
The culture, technology, processes and governance of data access for AI are likely outdated - appropriate access will need to happen much faster and at a greater scale
Sustaining society’s trust whilst also maximising your AI opportunity
Technology & continuous improvement is yet another unfunded indirect cost in the research sector!
Hence, the three near to middle-term recommendations actions are:
entrain to catch up
seed the missing expertise
underpin generative AI omics platforms
BioCommons, BPA's digital arm, already does (3) for the pre-AI era. It is clearly in the driver's seat to discover and operate the same for the AI era. BPA's partnership arm already coordinates the capture and reuse of data and is similarly placed to drive AI-era value creation through data. To quote Jenson (CEO of NVIDIA, perhaps a decade ago): if you don't build it they won't come! The only question for everyone is - do you have the evidence base to believe, and do you access to the capabilities to get there!?!