DDReg Pharma

Enhancing drug discovery processes through artificial intelligence

From pre-clinical phases to post-marketing follow up studies, the drug development process is a challenging, expensive, and time-consuming process that can take many years to complete. Companies spend millions of dollars and countless hours to test the effectiveness of their new drugs. To save cost and time, pharmaceutical companies are incorporating artificial intelligence and automation to make drug and molecule discovery processes more efficient.
A glance into the current drug development process
The drug development process includes steps from early research and development of a new drug to its market introduction. In the process, companies create a theoretical model of the desired drug’s clinical effects and determine how the drug will work in the body. Next, scientists try to reproduce the model in lab and animal studies. The process comes to an end once the drug has gone through all 3 clinical trial phases.
If the results are positive, regulatory approval is obtained for market introduction- more pharmaceutical experts and bodies are involved. For example, regulatory consulting companies provide pharmaceutical regulatory services to ensure the drug reaches the intended market; pharmacovigilance services are important to ensure compliance is met with global and/or regional standards at every stage of the product’s lifecycle. Risk management solutions provide procedures that companies must take to mitigate any risks that are associated with the drug. Labeling services such as product labeling and artwork compliance, and medical regulatory writing services are part of the entire regulatory process that is crucial for taking the drug to the market, all while maintaining compliance. Once the product is approved by relevant authorities, and has entered the market, companies must conduct regular post-marketing safety studies after the drug is on the market to ensure the risks are consistently and accurately monitored.
The drug development process is a meticulous one, involving research and development, regulatory filings, clinical trials and regulatory approval, market introduction, and market surveillance to ensure there are no safety concerns over the drug’s use.
New models for molecule discovery
There is no doubt that machine-learning models reduce the time and human effort required in complex processes. If early process can be automated and made more efficient, then the entire time to market for the product is reduced, allowing companies a larger return of investment. For example, in pre-clinical stages such as molecule discovery, machine learning models could suggest novel molecules and their characteristics that could help treat different conditions, in much shorter time. However, if these models suggest molecular structures that cannot be produced or replicated in the laboratory then this becomes challenging for companies as the molecule’s properties cannot be tested.
The two most crucial steps in molecular discovery are molecular design and synthesis planning. A study conducted at MIT proposed to combine the 2 steps thereby allowing a more streamlined approach to molecular discovery. This model uses networks to propose molecules from a pool of expert-curated templates and accessible compounds/materials that can be synthesized, and laboratory processes that can be followed. Additionally, this system suggests synthetic pathways in fractions of a second compared to other models that take several minutes.
The model receives a list of purchasable chemicals and reaction templates. This way researchers have control over the input that the model receives, which ultimately helps restrict the search space for the molecule. The model connects the inputs, where at each step the molecule becomes more complex. The final output is a refined structure of the molecule and associated chemical reactions used to synthesize it. This process also allows quality to be monitored at each level. Thousands of structures and chemical processes are fed into the model for it to learn from and automatically suggest synthetic pathways.
After receiving validation for the method, the team at MIT is focused on incorporating more refined reaction templates and enhancing the model as a whole; MIT’s model truly has the capability to change the landscape of molecule screening. By accelerating the molecule discovery phase significant time and cost will be saved, ultimately allowing an overall reduction in time-to-market for a pharmaceutical product- something every pharmaceutical company desires.
Authors: Ekta Kalra, Akshita Srivastava – Medical Content Team at DDReg Pharma
DDReg Pharma offers services in regulatory affairs, pharmacovigilance, medical writing, product labeling, publishing, and IPR support. We are your trusted partners for rapid market access and safety management.
Sources:
Gao, W., Mercado, R., Coley, C. W., (2021) Amortized Tree Generation for Bottom-Up Synthesis Planning and Synthesizable Molecular Design arXiv preprint arXiv:2110.06389

Leave a Comment

Your email address will not be published. Required fields are marked *

Send us a Message