Eli Lillyone of the world’s largest pharmaceutical companies, is leading a revolutionary change in the design of new medicines through the use of Generative Artificial Intelligence.
In recent experiments, this technology has proven capable of generating molecular structures at a speed and with a creativity that far exceeds human capabilitiesas shared Diogo Rauinformation and digital director of Eli Lillyin the Technology Executive Council Summit CNBC.
The AI is exploring millions of molecules, managing to identify drug candidates with “strange” structures, which do not correspond to anything known in the company’s existing molecular databases, but which present considerable potential. “As many molecules have been generated in five minutes as Lilly could synthesize in a whole year in conventional laboratories”Rau explained.
The possibility of AI generating medicines entirely on its own in the near future It will not only radically change the pharmaceutical industry, but also the scientific foundations that have been in place for centuries.
According to rauthe initial reaction of the executives of Lilly When faced with the designs generated by AI, there was skepticism and they even expected that scientists would find flaws in the proposals. However, the response from the researchers was surprisingly positive: “It’s interesting; “We had not thought about designing a molecule in that way”they indicated.
This openness to unusual designs marks a new era in pharmaceutical innovation, where Collaboration between humans and machines translates into mutual enrichment of creativity and scientific ability. “We always talk about training machines, but the art where machines produce ideas from a set of data that humans would not have been able to see or visualize stimulates creativity even more,” Rau reflected.
The medical revolution driven by artificial intelligence (AI) achieved a significant achievement with the creation of AlphaFold by the AI unit Google, DeepMind. This innovative approach marked a before and after in the field of biology, allowing a deeper understanding of the structure of proteins, crucial for the development and design of drugs.
Kimberly Powellvice president of healthcare at Nvidiahighlighted the ability of transformative AI models to interpret amino acid sequences and determine protein structure at unprecedented scales and resolutions.
This extraordinary achievement has translated into new strategies for the analysis and design of drugs, taking advantage of the vast catalog of chemical substances already digitized. According to Powellthis makes it easier for the AI to perform its training “in an unsupervised and self-supervised manner”, significantly expanding the capabilities possibilities of imagining drug models that would be beyond the reach of traditional human thought.
The revolutionary process does not stop there. The technology behind AlphaFold has been compared with the operational mechanisms of ChatGPT. As pointed out Powell to CNBC“has been trained on every book, web page, PDF document” and is able to encode the world’s knowledge in such a way that questions can be asked and answers generated.
The digitization of biology, at previously unimaginable scales and resolutions, forms the basis of this revolution. The inclusion of the spatial genomics allows the scanning of millions of cells within tissue in 3D, a technique that, together with the ability of AI models to work with chemicals in digital format, is marking the beginning of a new era in medicine.
The introduction of AI supercomputers and techniques inspired by the GPT model (Pretrained Generative Transformer) is allowing the simulation of the biological and chemical behavior of potential drugs, which could significantly accelerate development times and increase success rates in clinical trials.
According to recent studies, such as the one published in Nature by Amgenthis new approach not only reduces the discovery cycle from years to months, but also raises the probability of success from an uncertain 50% to a promising 90%.
Conventional drug discovery methods are often artisanal and expensive, with a failure rate close to 90%. This process involves extensive experimentation, human data analysis, and a design cycle based on the results obtained, followed by multiple decision stages in the hope of moving toward successful clinical trials.
An innovative aspect of AI in pharmaceutical research is its ability to generate new targets from existing data, which could reveal previously unknown therapeutic targets.
This ability of AI to explore new possibilities could break the trend of focusing on a limited set of drug targets, significantly broadening the horizon of research and development of new drugs.
The reduction in the time and cost of developing new drugs It is a crucial benefit of the implementation of the AI in this field. With costs ranging from $30 million to $300 million per clinical trial, the ability to shorten development timelines represents not only a scientific advance, but also a significant economic opportunity for the pharmaceutical industry.
