Generic artificial intelligence (generative AI) is now being used to transform scientists’ instructions for molecules with specific characteristics into new drugs for diseases like cancer, Alzheimer’s, arthritis, fibrosis, and other rare diseases. This is similar to how ChatGPT transforms text directions into new written materials or how DALL-E 2 produces realistic-looking images from prompts.
Through the use of generative models, transformers, and other cutting-edge machine learning (ML) techniques, AI can be trained to sift through enormous, complex stores of biological and chemical data to discover novel therapeutic targets and create novel molecular structures with desired properties.
For the treatment of cancer and other diseases, Insilico Medicine’s generative AI drug design platform, Pharma.AI, has created several AI-designed medications. These include a USP1 inhibitor for the treatment of solid tumors that has FDA approval for clinical trials and a QPCTL inhibitor that may be a first-in-class inhibitor for malignant tumors that is being developed with Fosun Pharma. Phase II patient trials for the company’s main therapy, which treats the uncommon progressive lung condition known as idiopathic pulmonary fibrosis, have been reached—a significant development for a fully generative AI medication.
AI is also being used by other businesses to expedite innovative therapies. For rare diseases like cerebral cavernous malformation, which causes enlarged and atypical blood vessels in the brain, and neurofibromatosis type 2, a genetic condition that results in benign tumors of the brain and spinal cord, Recursion is using in-house data generation and cutting-edge computational tools to develop new treatments. These two programs are currently through phase II clinical studies. Healx, a business, uses artificial intelligence to help find existing medications that can be successfully repurposed to treat rare disorders. It is working on treatments for cancers like plexiform neurofibroma and rare disorders like fragile X syndrome, which include tumors that arise from nerve cells.
Even though computational techniques for small molecule drug design have been around since at least the 1970s, according to Nature, it has only been recently that researchers have been able to employ AI to screen datasets for interesting targets and create new medications.” ML algorithms have progressively developed into a deep learning technique with potent generalization capability and more effective big data handling,” according to a review in the journal Molecular Diversity. “This further promotes the integration of AI technology and computer-assisted drug discovery technology, thereby accelerating the design and discovery of the newest drugs.”
According to the National Human Genome Research Institute, scientists have now accumulated significant databases, including “extraordinary amounts of genomics data.” The institution estimates that within the next ten years, genomics research will produce between 2 and 40 exabytes of data, which will be further multiplied and complicated by DNA sequencing and other biological technologies.
As stated by the organization, “Genomics researchers need AI/ML-based computational tools that can handle, extract, and interpret the valuable information hidden within this large trove of data.”
Generative AI in Medicine: Moving Forward
A young computer scientist named Ian Goodfellow’s key 2014 study, Generative Adversarial Networks, is where generative AI first made its appearance. He learned that technology may be used to process massive volumes of data to create fresh, synthetic data with particular properties. The quality of the synthetic data improves with increasing data and feedback.
Now that the same technology has been used to create a variety of new materials based on data—from stories to art—the potential for discovering novel molecules to treat disease might have profound effects.
Traditional drug discovery is a relatively expensive, ineffective, and long procedure. The process of bringing a medication from its invention to commercialization takes 10–15 years and costs billions of dollars. And those expenses keep going up.
Every step of the process is accelerated and more effective thanks to generative AI, which also helps predict the likelihood of clinical trials, where the majority of drugs in development fail, and design novel drugs specifically suited to act on those targets. InClinico, a generative AI tool under development, is intended to forecast the likelihood that a clinical trial will be successful at the crucial transitional point between phases two and three. According to a recent study published in Clinical Pharmacology and Therapeutics, after a seven-year study, it showed 79% accuracy in predicting the results of real-world trials.
Additionally, AI is being developed to produce synthetic DNA that may aid in the creation of more effective vaccines, as well as digital twins for clinical trials to help with patient screening and diagnosis.
According to Associate Professor Aleksej Zelezniak in Drug Target Review, “Now we have succeeded in designing our DNA that contains the precise instructions to control the quantity of a specific protein.”
Early in my career, while working as a member of a well-known team of scientists searching for genes linked to pediatric cancer, I was personally introduced to the transformative ability of data and machine learning to hasten my understanding of disease and the creation of novel therapies. I was in charge of utilizing a cutting-edge deep-learning technique to reduce the gene expression dataset’s dimensionality. Putting on 3-D glasses to investigate the visualization of patient data was a powerful epiphany regarding the potential of this cutting-edge technology. The first generative AI medications are now being developed by firms, and several are currently undergoing human clinical trials.
It’s a crucial time. Although the first AI-designed medications have not yet hit the market, it is encouraging to see more collaborations between biotechs and well-established pharma, such as those between BenevolentAI and AstraZeneca, Insilico Medicine and Sanofi, and Exscientia and Bayer, as well as numerous pharma companies developing their own internal AI capabilities, including GSK, Novartis, and Roche.
Only four significant pharma partnerships for AI-based drug discovery were formed in 2015, according to research from GlobalData. There were 27, a 575 percent growth by 2020. The increasing tendency is still present.
The future of medicine could be drastically changed if deep, high-quality data, knowledge, and increasingly complex AI capabilities are combined.
