Founder and CEO of Colossal, Hypergiant, Conversable, ChaoticMoon and TeamChaos.
We are at the infancy of the biotech industry’s adoption of machine learning tools. But, over the next 20 years, a more “multidisciplinary and data-intensive approach to life sciences will shift our understanding of and ability to manipulate living matter.” However, lack of data sophistication is limiting the potential for rapid advancement right now. And, in my opinion, the more we wait, the more we will fall behind.
Failures to adapt in biotech are largely due to the challenges of data collection, information silos and the fact that historically, the industry is biology-first rather than software-first. Yet, when tackling systems-level challenges, as biotech does, machine learning adoption is unavoidable to achieve scale and cost-efficiency over time. Business leaders who do not understand this will fall far behind those who are building biotechnology companies with data science at their core.
Biotechnology companies need to adopt modern data practices.
Across biotech, there is a growing amount of uncategorized and scattered data. It is projected that in the coming years, exabytes of data will be generated by the work within biological sciences. This is due to the large number of experiments required to prove concepts. Any time people are following the same type of experiment or analysis more than a handful of times, software is widely accepted as more efficient than a human. Yet, the industry does not have mature data collection and management standards necessary. This needs to change.
The more software the industry utilizes, the more it can collect data. That data can then be analyzed with machine learning, and new inferences can be made. Those inferences may not have been the intent of the experiment, or even an obvious finding given the many sets of experiments humans may have had to review at once historically. Concurrently, the ability to create hypothesis-free tests means that there is also the chance to generate scientific insights that might never have been uncovered otherwise. That’s a lot of data.
At my company, experiments can cost millions of dollars, but we need tens of thousands of experiments to bring scale to our efforts. This is true whether biotech companies are attempting to de-extinct species, create personalized healthcare or run drug discovery trials. To reduce cost and effort, we must augment biological trials with machine learning workflows, including software running complex data sets.
The ability for machine learning to study fundamental aspects and relationships of biotechnology data and bring to life new concepts is where its power and advantage sits. Importantly, we’re seeing the software industry begin to support these efforts. Take Nvidia for example, which launched its BioNeMo Large Language Model (LLM) service to “help researchers build new artificial intelligence (AI) models for biology and it’s an effort that has yielded some strong early results.” They used BioNeMo with Evozyne to create the Prot-VAE model for protein discovery in just a few weeks.
Modern data practices can usher in an era of machine learning.
Biotechnology companies can adopt software strategies to improve their capacity for machine learning today. Here are three tested applications I have utilized that may be relevant to the work of others:
1. Genome Editing For Innovation
Utilizing machine learning with large sets of biological data has allowed my company to find similarities in how genes express traits. By looking across species, scientists are able to find out more about how body size, morphology, physiology, color/patterning and protein are determined. To do this, we connected our phenotype research team who engages in genomic analysis for how genes are conducted as traits and with our bio-informatics engineering team. Together they were able to build a unique tooling for phenotype expression research, all on Google Cloud. Then, our proprietary machine learning tools were able to analyze that data and decode phenotype traits in species.
2. Hardware For Automation
Further, Colossal’s Ex-utero Development team has collaborated with our machine learning team to develop algorithms that automate the tracking and quality assessment of cells. By doing this, you can scale your throughput of analysis and continuously find the most healthy and viable cells to work with. This has resulted in the automating tracking of cellular development, growth and health.
3. Field Experiments Related To Conservation
Finally, my team leverages machine learning solutions in addition to building our own, in order to apply it to additional conservation work—a field in which there is slow adoption of any emerging technology. We consider it a low technology transfer field. Even using applications like computer vision to monitor animal populations or audio analysis to track passive acoustic monitoring of animals has rarely been done at scale. The data gathered allows us to better understand how species act, think and grow.
The bottom line is that urging biotechnology companies to adopt mature data collection and management standards does not mean that they need to also create cutting-edge data solutions. It is possible to adopt industry best practices in software and apply them to many aspects of the field of biological sciences.
But, a failure to adopt the most urgent and promising technologies means that the solutions we need to address the continued onslaught of disease, changing climate and health conditions will not be met within our lifetimes. That is simply unacceptable. It is a problem we as leaders have the existing solutions to fix.
Now, enterprise decision-makers must encourage, fund and build enterprise-wide machine learning solutions within biotechnology. This means prioritizing the adoption of technology advancements in budgeting, within traditional scientific teams and across the IT infrastructure. To do this, biotech companies must elevate computer scientists to the same levels of seniority as other leaders and ensure they are given the remit to see and create collaborative learning environments. This can help rapidly transition companies forward and ensure that biotech truly meets the demands and opportunities of the next twenty years.
Forbes Business Council is the foremost growth and networking organization for business owners and leaders. Do I qualify?
Read the full article here