Banner by Alisa Zhang

AI Driven Drug Discovery

By Jayashree Ezhilarasu

Would you take drugs discovered and designed by artificial intelligence? Clinical drugs can take anywhere from six to seven years to be established, and only one out of ten of these drugs achieves clearance after promising trials, resulting in money and time being lost. Drug discovery refers to the process by which new therapeutic methods are found to treat illness. This process can be expensive and timely. Artificial Intelligence (AI) has been implemented progressively by many biotech companies, serving as a means of expediting the drug discovery process and conserving time and resources. 

Artificial intelligence (AI) is technology that shows intelligence similar to that of humans and it works by applying learning algorithms to analyze data. AI can consume a large amount of data and predict the efficacy of clinical drugs while identifying potential side effects, thus promoting safety. AI can also quickly produce the documentation, reports, and data needed throughout the development process, and it does so with few mistakes and consistent language. Another use is its ability to predict the other target interactions a drug can have with an individual, possibly allowing certain drugs to be relaunched for phase 2 trials for certain conditions. According to the The Food and Drug Administration (FDA), there are five stages to clinical drug development: discovery and development, preclinical research, clinical research, FDA drug review, and FDA post-market drug safety monitoring. Within clinical research there are four phases: phase 1 is to establish safety and dosage, phase 2 for efficacy and side effects, phase 3 for efficacy and monitoring adverse reactions, and phase 4 further checks for safety and efficacy after the drug is approved by the FDA. The number of participants and duration of these phases vary with phase 1 lasting a few months with 20 to 100 healthy volunteers or people with the condition and phase 3 lasting one to four years with 30 to 3,000 people with the condition. Relaunching a drug directly into phase 2 of clinical trials is much cheaper than starting trials for a completely new drug ($8.4 million compared to $41.3 million).

The Food and Drug Administration (FDA) granted its first orphan drug status to an AI-designed drug in February 2023, designed by Insilico Medicine. According to the National Institute of Health (NIH), an orphan drug designation is given to drugs “that show promise in the treatment, prevention, or diagnosis of orphan diseases.” Orphan diseases refer to a condition that afflicts fewer than 200,000 people, or, in other words, a rare disease. INS018_055 is the first AI-designed drug to enter phase II clinical trials in humans as of June 2023. This drug was created using generative AI to treat idiopathic pulmonary fibrosis (IPF), a rare lung disease. This is a major step for clinical drug development as well as treating rare diseases, as the process of finding therapeutic drugs is simplified through AI technology.

A significant contribution to the field of technology is Open AI’s ChatGPT, which is a large language model (LLM). ChatGPT is trained on data texts from the internet and simulates human interaction with easy-to-digest information. In March 2023, Insilico announced the release of its target discovery chat platform, ChatPandaGPT. Insilico Medicine has made their data more accessible by implementing a chat function, allowing users to converse and inquire about their data. ChatGPT moderates the chat output; however, Insilico utilizes its own AI to verify the information produced.

 Moreover, Insilico Medicine is the first biotech company to integrate a large language model (LLM) with an AI drug discovery platform. Omics informally refers to the several branches in biology ending in the suffix -omics such as genomics, proteomics, and metabolomics. Their platform, PandaOmics, utilizes its stored OMICs datasets to apply its artificial intelligence hypothesis generation system to rank and evaluate target genes for a disease of interest. It also pulls any connections from publications, grants, patents, or clinical trials. It can then generate background information regarding the target of interest and predict its potential to enter phase 1 clinical trials. Moreover, PandaOmics can estimate the growth of attention to a gene over a five-year period and predict the attention it will receive in the next three years. One of the major uses of PandaOmics is its graph analytics platform, which visually shows the links between genes, diseases, processes, and other compounds. Insilico Medicine integrates the chat feature in order for researchers to have a more natural conversation with the platform and browse through data in a more efficient manner.

ChatPandaGPT provides thorough information regarding the characteristics of a gene or disease, including gene-disease interactions, major signaling pathways, and potential side effects of certain drugs or prior clinical trials. The user can pick from predefined questions or create their own. Some examples of queries listed on the Insilico website are: “How does PLK1 link with apoptotic processes in sarcoma?", "Are there any targets for monoclonal antibodies for sarcoma?" "What diseases can be treated by alvocidib?", "Besides EGFR, what are the other potential targets of gefitinib?" Overall, the usage of AI in drug discovery and development saves researchers time, money, and resources.

The integration of artificial intelligence (AI) in drug discovery and development has revolutionized the field by accelerating the process and conserving valuable time and resources. By combining AI technologies with drug discovery platforms, researchers can streamline the development process, improve success rates, and ultimately bring life-saving treatments to patients more efficiently, especially those with more rare diseases. AI is transforming the future of clinical drug discovery, putting forth great strides for the future of health care.