Over the past several years, artificial intelligence (AI) technology has been used across drug development and discovery to help bring novel medicines to market more quickly and cost-efficiently.
The rise of data digitization, in addition to advances in automation and AI technologies, have given researchers in drug development new pathways to leverage AI-based models to lower costs, expedite discovery, and even improve patient outcomes.
And because of the complexity and sheer volume of available data today, there’s a growing need for additional AI and machine learning models which can parse through these vast quantities of data and find meaningful patterns.
Now, researchers are beginning to look at the utility of generative AI in drug development for specific ways this particular type of machine learning technology might help streamline, improve, or otherwise accelerate the development of new drugs.
Here’s what you need to know about generative AI and drug development.
Generative artificial intelligence describes a broad category of machine learning algorithms that recognize patterns and structures within a set of data to generate original content.
To do this, generative AI models use a type of deep learning called generative adversarial networks (GANs). GANs consist of two neural networks: a generator that creates new data and a discriminator that evaluates the data. These work together to iteratively improve the output of the model. While traditional AI systems are also designed to recognize patterns, they typically do so to make predictions. Distinctively, generative AI creates new content from its learnings – precisely the reason these models are called “generative.”
Large language models (LLMs), which have gained particular prominence in recent months with the launch of OpenAI’s ChatGPT and Google’s Bard, represent a type of generative AI which can create human-sounding language and generate never before seen combinations of text. Other generative AI models use data patterns to create different types of content like code, images, text, music, and video.
AI-based technology is already being used in drug development – for example in automating biomarker discovery, in analyzing large-scale omics and clinical datasets for patient stratification, and in first-in-human dose optimization. The benefits of these advancements are plentiful – from helping diagnose diseases earlier, to mitigating disease progression, to accelerating the development of more effective treatments.
Generative AI also has enormous potential to help revolutionize many aspects of drug discovery and development. Trained on vast amounts of data, with the ability to virtually synthesize images, text, speech, and image captions, medical researchers are exploring ways to leverage powerful generative-based models to expedite drug discovery and design new drug and biologic molecules. These algorithms can learn from chemical datasets how to predict structures of de novo proteins and molecules with desirable properties, or predict the 3D binding structure of a small molecule. All of these deep learning methods can ultimately help facilitate the engineering of more effective candidate drugs.
Some in the generative AI space are already using generative AI in order to create synthetic data for research purposes, particularly in areas like rare disease, where lack of patient data can be a challenge. Though these methods are still quite new, they hold real potential for the industry and patient care, in general.
In addition, generative AI can be useful in automating many of the tasks related to drug discovery and development, for example, coding, imaging, scribing/documentation and even genomic analysis. New LLMs trained on biology and chemistry texts may be able to help researchers better understand the behavior of proteins, small molecules, DNA, and a range of other biomedical phenomena.
Generative AI-based drug discovery isn’t meant to replace human intelligence. The goal is to support the work of scientists so they can do their jobs better and with more efficiency, potentially developing game-changing, life-saving insights along the way.
Some still might wonder how to trust the safety of a molecule designed by generative AI. To help build confidence in the technology, the integration of explainable AI (XAI) capabilities can help validate many of a model’s processes, improve transparency, and provide explanations for how machine learning models work.
Hallmarks of XAI include: descriptions of how predictions or classifications were made, a mode for debugging questionable results or behaviors that occurred during modeling, and regulating features that help avoid bias.
Broader applications for generative AI in healthcare are being explored, particularly as the available body of medical literature and patient health data expands. Personalized medicine, which promises to use an individual’s own genetic data to prevent, diagnose, or treat disease, is one area to watch. As generative AI technology improves, it may have a role across a range of operational processes in drug development, for example patient scheduling and flow of clinical trials.
VeriSIM Life has developed its own sophisticated computational platform that leverages advanced AI and ML techniques to improve drug discovery and development by greatly reducing the time and money it takes to bring a drug to market. Contact us to learn more about BIOiSIM™ and how our AI-enabled platform helps de-risk R&D decisions.