Artificial Intelligence (AI) technologies are currently the most disruptive trend in the pharmaceutical industry. Over the past year, we have quite extensively covered the impact these intelligent technologies can have on conventional drug discovery and development processes. This series of blogs summarizes our key takeaways from this research, along with practical advice for both early- and late-stage drug developers who are looking to leverage AI in their research programs.
We charted how AI and machine learning technologies came to be a core component of drug discovery and development, and their potential to exponentially scale and autonomy this function. AI has been integrated into every step of the drug discovery process from screening data, models and simulations. It is able to expand the scope of drug research even in geographically and data-limited specialities like rare diseases. The power of knowledge graph-based drug discovery will transform a range of drug discovery tasks including target validation, virtual screening through to scaffold evolution, ligand design and synthesis evaluation.
AI is the new frontier of drug discovery and development. AI technologies can radically remake every stage of the drug discovery and development process, from research to clinical trials. Today we’ll delve a bit deeper into the transformational possibilities of these technologies in two foundational stages — Early Drug Discovery and Preclinical Development — of the drug development process.
Early Drug Discovery and Preclinical Development
The early stages of drug development are a critical and complex process that determines the profitability and value of downstream research. This brief white paper explores the challenges that exist in this stage, as well as novel solutions for making incremental improvements in accuracy and efficiency so that you can maximize your ROI at all stages of drug development.
AI/ML in early drug discovery
Drug discovery is a process that identifies and uses chemical compounds that can lead to the creation of new medicine. The early drug discovery process flows, broadly, across target identification, lead identification, lead optimization and finally, candidate selection. Currently, this time-consuming and resource-intensive process has a high reliance on translational approaches and assumptions.
The process of creating new drugs is an expensive and time-consuming endeavour. And while the first stage of development indicates whether the drug is safe and well-tolerated, it does not reveal whether or not it is effective. As a result, a new molecular entity (NME) may well enter late-stage development without adequate evidence that it will meet quality or performance expectations. Investigating drug-target interactions (DTIs), therefore, is a critical step to enhancing the success rate of new drug discovery.
Predicting drug-target interactions
Drug discovery is still an expensive and time-consuming process. Drug discovery takes a lot of resources and money to develop a new drug. Drug developers need to screen millions of molecules in order to find one with the right properties (chemical, biological and pharmacological).
We introduce a new approach for Multi-DTI models that balance bias and variance through a multi-task learning framework. In the new approach, both tasks are performed with similar feature representations of CNNs with a co-attention mechanism. The resulting model shows promising performance on the publicly released Multi-DTI benchmarks.
Lead identification & optimization
This stage focuses on identifying drug-like small molecules that exhibit therapeutic activity and optimising these potential molecules. The challenge in this hit-to-lead generation phase is twofold. First, the search space to extract hit molecules from compound libraries extends to millions of molecules. And second, the hit rate of conventional HTS (high-throughput screening) approaches to yield a viable compound is just around 0.1 %.the preclinical study is a stage of drug development. In this stage of the process, researchers test drugs in living organisms. At this point, there are still no people involved in the early studies.