AI is not the Silver Bullet

• Read original article here

As innovators, entrepreneurs, and operators, we have a responsibility to ensure that  our companies, teams, services and tools are successful. With that said, oftentimes there is a tendency to put part of that responsibility on the shoulders of our tools when we deal with Artificial Intelligence (AI) systems. We might see a day to have conscious computer systems, but it is not today, and we are still responsible for developing fair, unbiased, reproducible and successful AI technologies. As an industry, I feel strongly that when we develop an AI tool, release it publicly or sell it as a service, it behooves us to take that responsibility seriously. 

An important philosophical foundation at Cyclica is that AI is not the silver bullet, and that a deep empathy towards the complexities of biology, genetics, and chemistry is required to make a meaningful impact in the field of drug discovery. In other words, we still need Human Intelligence (HI) in this field. We believe this mindset in developing and using AI technologies has been a determining factor in our accomplishments to date.  

Data-driven and AI-based drug discovery can be a boon to health equity, as a means to speed up and reduce the cost of medical research and development. The more data, the more inputs, the more connectivity and cooperation that is possible in the life sciences field, the more breakthroughs we can have.

There’s always so much more work to do, and fighting to get our hands on more, higher quality, balanced, and fair data is important. We do this by deeply evaluating public and acquired datasets prior to integration, including both positive and negative data back into our models where appropriate, and partnering with leading institutions globally to integrate their vetted data. With so many millions of proteins and molecules at researchers’ disposal, it would be an impossibly time-consuming task without AI. And that AI is itself so large and complex, computationally speaking, that it can only be hosted on the cloud to be effectively accessed anywhere and any time. 

Cyclica was one of the first companies that integrated the AlphaFold2 predicted protein structure data released by Google-affiliated DeepMind. In July 2021, researchers working at the company and a number of European scientific institutions were able to computationally map the millions of protein variations with the same confidence achieved in an experimental setting. 

Through Google Cloud, Cyclica was able to immediately adapt its proteome pipeline to work with its newly predicted structures (here). Currently this database encompasses more than 500,000 proteins across 48 species and comprises more than 5,000,000 potential pockets. This represents the industry’s largest known computational-powered human proteome and enables Cyclica to efficiently address proteins of non-human species like bacteria, viruses, fungi, and rodents. You can learn more about our work and perspective on AlphaFold2 in a running blog series here. 

Cyclica has been developing and improving successful AI technologies relying on the human intelligence and tremendous efforts of Cyclicans. The day has not arrived, yet, where a machine can train a machine to create a drug. The level of futuristic thinking is inspiring to many, but we need to ground ourselves to the truth: that AI is not the silver bullet, nor is HI. They both have to work in unison for us to make a demonstrable impact on how drugs are discovered and brought to patients faster and more equitably.