Have you ever heard of Ibuprofen? Chances are, yes, as it is among the most commonly prescribed drugs in the world. You have perhaps even used it to treat pain or fever. What you may not know is that developing and optimizing such medications, making sure they are effective and safe, is a complex and time-consuming process.
As consumers, it is easy to underestimate the resources which are dedicated to drug development. The process (see Figure below) can last as long as a decade, and cost several billions of dollars. The success rate for drug development (from the identification of a drug candidate to drug approvals) is very low: a recent study [1] estimated it at 6.2%.
Therefore, the use of Machine Learning (ML) methods is driven by a need to lower the overall cost and complexity of the pipeline. Fortunately, the images, textual information, and biometrics collected at various stages of drug discovery and development by pharmaceutical companies are tasty nourishment for data-hungry ML algorithms!
The wide availability of GPUs has enabled remarkable progress to be achieved in the field of artificial intelligence. However, we believe that photonic computing, along with the development of new ML algorithms, can transfigure the current applications of ML within pharmaceutical companies.
Let us explore some examples of how the use of photonic computing could transform drug design.
The first step in drug discovery is to identify a molecular target (gene or protein) which, if activated or inhibited, affects a given disease. We already possess most of this knowledge: the biomedical literature is full of information on how a certain target is associated with a certain disease. However, the increasing size of scientific literature calls for an automated process to unlock the relevant information.
Photonic computing could be a key for the next generation GPT-X. Such an explosive growth in natural language processing will allow a much more efficient automated processing of the available literature. Extracting relevant information for target identification will become seamless.
Once a compound that can act on a target molecule has been identified, it’s time for synthesis! To plan a route for organic synthesis, we employ retrosynthetic analysis: the compound is formally decomposed into simpler precursor structures using reversed reactions, and each precursor is analyzed using the same procedure. This results in a sequence of reactions to synthesize our final compound from simple or commercially available structures.
