In recent years talk of AI (artificial intelligence) (1,2) has picked up speed and progressively been touted as the methodology/technology to go to for new products, manufacturing process technology development, quality, supply chain, and inventory management etc. for the active pharmaceutical ingredients (API) and fine/specialty chemicals. When fully developed and incorporated, it is suggested and expected to be the salvation of almost every current shortcoming. I have not heard much about the use of AI based applications in the pharmaceutical products and fine specialty chemicals, if it is happening.
This post is an update on my last post on the subject (3). There is no financial relationship with any “for profit” and non-profit” organization. We all know that API and fine/specialty chemicals are organic chemicals that have similar skeletal structures and manufacturing processes. The only difference is that one set enhance life style and the other extends/prolongs life. Many are suggesting value of AI in the new pharmaceuticals discovery and/or their manufacturing. I wonder how much of own monies has been invested or spent in R&D and manufacturing by who are suggesting the perceived/resulting benefits. Are there verifiable examples? If any of the pharmaceutical companies are exploring AI applications, they are not being discussed in open forums.
Application/inclusion of AI in pharmaceuticals is possible and it can happen by taking one step at a time. Since the pharmaceutical manufacturing is regulated, my conjecture is that what all is being practiced today and is in the pipeline will not change for the brand as well as the generic drugs. Related expense is too high and no business is going to absorb it and/or pass it on to the patients. Potential application of AI to process quality and manufacturing process control is suggestive that our universities have done an inadequate job of training process designers. Every AI proposed manufacturing specific application will have to be ROI justified and tested especially if it impacts manufacturing. Regulatory approval could be a deterrent.
New Product/Drug Identification:
AI based new drug product scouting for a disease is a possibility. Someone has to pay for the effort. No one knows the time or the investment AI route will take. Time and expense for AI based effort has to be lower that the current effort. Someone will have to take a leap of faith. If it takes same as the current time and money, AI based effort will fizzle out.
Product Manufacturing Process:
Once a potential product is suggested/selected by AI or a current method, the real task of taking the identified product to commercialization begins. If the companies stay with their current practices and methods (4) nothing will change and companies will not achieve Net ZERO (5). AI recommended methods will have to be better and economic.
I thought it would be a great idea to ask ChatGPT (1) what does it know about sociochemicology (5,6,7). It could not give any answer. On sharing a video link (6) that briefly explains sociochemicology, it was very quick to understand and share its value in development of efficient chemical processes and achieving “Net Zero (5) ” in pharmaceutical and specialty chemicals. Its search capabilities are impressive.
I decided to check out what AI can suggest for process chemistry and manufacturing routes for few chemicals I am very familiar with. I tested ChatGPT (1) and Perplexity (2) by asking my credentials. “Results were interesting. They accurately identified few things but could not give details. I was told by others that AI it is constantly learning and it becomes better by answering questions. I asked additional questions and it cited some of my work.
I queried ChatGPT (1) about the following chemicals.
Saccharin:
I asked ChatGPT (1) about the manufacture of saccharin. It gave me a process practiced by Monsanto which was abandoned in late sixties. When reminded that there is an alternate process starting with phthalic anhydride, it gave me its process details. It was significantly different and complex chemistry compared to what we practiced which was phthalic anhydride based also. It surprised me as I was involved with scale-up and manufacture of our chemistry, a continuous process that operated year round with scheduled shut downs for maintenance etc..
Omeprazole:
I asked about Omeprazole chemistry. ChatGPT (1) does not suggest how to execute the its chemistry. Chemistry outlined in USP 7,227,024 (8,9) is an alternate process. Compared to ChatGPT (1) suggested process ‘024 based process’s execution is very simple and can be produced using a continuous process. It includes some of the sociochemicology (6,7) based teachings.
Metformin Hydrochloride:
ChatGPT (1) suggests a process chemistry. It does not suggest its manufacturing process. A quick review of this chemistry suggests it can be a challenge to commercialize. An alternate chemistry and method (10, 12) suggest a safer continuous manufacturing processing.
AI suggested chemistries for the above three products are not economic as better and economic processes were/are commercial. Use of sociochemicology (6,7) is profusely incorporated in each of the above actual and suggested processes (8,9,10,12) .
Future Path:
Considering pharma’s current commercialization and manufacturing practices, about 60+ years old (10, 11, 12), each of the following that could be AI based will have to be considered and evaluated for their economic and commercial merit.
· If the drug discovery is going to take same/similar time like the current time and path, my speculation is that all the euphoria and the related investment in AI could be lost. It can provide chemistries for the identified product but I am not sure it can generate and economic processes. This observation is based on the three chemistries I inquired about. Would AI be able to suggest the most optimum chemistry and manufacturing processes for the new molecules is not known? Even if it does, each would need a thorough review and it will have to be tested.
· For inclusion of AI each company will have to decide its commercial details e.g. business model, economics, investment, operating strategies etc.. One would expect that village’s (chemists, chemical engineers, purchasing, maintenance, accounting) journey to “net Zero” (6, 12,13) would begin when a chemical molecule is identified and/or speculated as a potential drug candidate. If a pharma company decides to fit the process in its available equipment (4), its manufacturing methods will stay 60+ years old (10, 11, 12, 13) and it will not achieve “net zero (5)”.
· If companies want economic “net zero” (5) processes they will have to exploit mutual behavior of chemicals, sociochemicology (6, 7) and Real Information/Intelligence (RI) (5, 6, 7, 10, 11, 12, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23) and exploit capabilities of the processing equipment that are available and being used in other manufacturing industries (5). It is interesting to note that each process chemistry tells us about the chemicals used and produced. Their sociochemicological behavior can be capitalize on to create excellent manufacturing processes (10, 11, 12, 23). It seems somehow pharma industry has not been capitalizing on this information. Their inclusion and consideration along with use of modular processes (10, 11, 12, 23) can deliver excellent Net Zero (5) processes.
· Many are suggesting application of AI for e.g. inventory control, quality control, manufacturing efficiencies, process control and supply chain etc. in pharma. It is difficult to postulate the impact of new applications. I don’t know whether if any of the AI based application suggesters have invested their own money to test their applications in pharma’s current manufacturing operations and received regulatory approval. Considering regulatory challenges this could be hard.
If pharma companies decide to bring their manufacturing from antiquity to today and move forward, AI might be of value. However, each company will have to decide their incorporation based on cost, value and regulatory interference. Let us see where does AI lead pharma in the next two years. It would be worth re-visiting the topic.
Girish Malhotra, PE
EPCOT International
1. ChatGPT https://chatgpt.com
2. Perplexity https://www.perplexity.ai
3. Malhotra, Girish: Artificial intelligence in Chemical Process Development, Manufacturing & Net Zero, Profitability through Simplicity, March 31, 2023
4. Malhotra, Girish: Square Plug In A Round Hole: Does This Scenario Exist in Pharmaceuticals? Profitability through Simplicity, August 17, 2010
5. Malhotra, Girish: NET ZERO for Active Pharmaceutical Ingredient & Fine/Specialty Chemicals: Nondestructive Creation, Profitability through Simplicity, November 7, 2024
6. Malhotra, Girish: Sociochemicology , May 30, 2013
7. Malhotra, Girish: Focus on Physical Properties To Improve Processes: Chemical Engineering, Vol. 119 No. 4 April 2012, pgs. 63-66
8. Malhotra, Girish: Alphabet Shuffle: Moving From QbA to QbD - An Example of Continuous Processing, Pharmaceutical Processing, February 2009 pg. 12-13
9. Malhotra, Girish: Analysis of API (Omeprazole): My perspective, Poster Session: Pharmaceutical AIChE Annual Meeting, November 11, 2009, Nashville, TN
10. Malhotra, Girish: Chemical Process Simplification: Improving Productivity and Sustainability John Wiley & Sons, February 2011
11. Malhotra, Girish: Chapter 4 “Simplified Process Development and Commercialization” in “ Quality by Design-Putting Theory into Practice” co-published by Parenteral Drug Association and DHI Publishing© February 2011
12. Malhotra, Girish: Active Pharmaceutical Ingredient Manufacturing: Nondestructive Creation De Gruyter April 2022
13. Schrader, Ulf: McKinsey & Co. Operations can launch blockbuster in pharma, February 16, 2021
14. Malhotra, Girish: Profitability through Simplicity
15. Malhotra, Girish: Focus on Physical Properties To Improve Processes: Chemical Engineering, Vol. 119 No. 4 April 2012, pgs. 63-66
16. Shreve, R. N. Unit Processes in Chemical Engineering, Industrial and Engineering Chemistry,1954, 46, 4, pg., 672, Accessed June 22, 2020.
17. McCabe W. L & Smith J. M. Unit Operations of Chemical Engineering McGraw-Hill Book Company Second Edition 1967
18. McGraw Hill Chemical engineering Series https://www.book-info.com/series/McGraw-Hill+chemical+engineering+series.mobi.htm
19. Malhotra, Girish: Process Simplification and Net Zero: Capitalizing on Physical and Chemical Properties of Reactants and Intermediate, Profitability through Simplicity, August 20, 2024
20. Gasteiger J.: Chemistry in Times of Artificial Intelligence ChemPhysChem 2020, 21, 2233– 2242 Accessed March 21, 2023
21. Perry, J. H. et.al. Chemical Engineer’s Handbook Fourth Edition: McGraw-Hill Chemical Engineering Series, 1963
22. Levenspiel, O: Chemical Reaction Engineering, John Wiley & Sons 1999
23. Malhotra, Girish: Process Simplification: Is the Best Answer: New Terminologies or the Application of Fundamentals? Profitability through Simplicity, December 14, 2024