Artificial intelligence in Chemical Process Development, Manufacturing & Net Zero

Artificial Intelligence (AI) ⁽¹⁾ is becoming the latest go to technology/methodology for almost every application. Numerous articles have been published in the areas of chemical properties, discoveries, process design and development and manufacturing areas ^{(2,3,4,5,6,7)}. If the machine language can be used to simplify conventional chemical process development and design, it is my speculation that commercialization of products can be sped up multifold. Chemical manufacturing landscape could change dramatically. This will lead to improved profits through faster availability and reduced shortages of products in every segment of the chemical industry that includes fine/specialty chemicals, additives, cosmetics, flavors and fragrances and pharmaceuticals. In addition, every chemical related business can make significant progress towards “net zero” ^(8,9) emissions.   

 

I am presenting my perspective on how AI could be used to develop, simplify, speed up development, commercialization and manufacturing of the products. Perspective presented here is my own. There is no financial relationship with any “for profit” and nonprofit organization.

 

AI and Process Development:

 

For most of the product categories mentioned above development and commercialization take their own course. This is due to availability of data such as properties and mutual behavior of the chemicals needed and used in product evaluation, formulation and process design. Development moves from the laboratory to commercial scale following their established paths that have been taught and practiced over the last 70+ years. 

 

For product development physical and chemical properties of chemicals are needed. Much of the data about chemicals (physical and chemical properties, safety and health value of the chemicals) used and produced might be available from Chemical Abstract Services ⁽¹⁰⁾ NIST Chemistry Web Book ⁽¹¹⁾, product suppliers and can be generated internally at the companies. AI should accelerate and facilitate availability of solubility and safety data. Safety data and solubilities of solvents identified by AI would have to be verified in the laboratory. It could also be used to define process conditions for chemical processing. Process development and commercialization will be accelerated.

 

Many of the chemicals are raw materials are solid at ambient temperatures. To facilitate processing and ease of handling, they are solubilized or slurried in inert solvent/s. These are recovered and reused. This has been the tradition for the last century. Even though the solvents are recovered and reused, they still leave a large environmental footprint than a process which is all liquid process i.e. solids do not have to be slurried to facilitate processing. 

 

Sociochemicology ^{(12, 13)} i.e. mutual behavior of chemicals plays a significant part in development and commercialization of chemical synthesis processes. Some of the methods that can be used to facilitate incorporation of physical properties of the chemicals are reviewed ^(14,15). However, the methods should not be limited to what has been discussed in literature. 

 

Some of the methods that value use of liquids in chemical processing are reviewed ⁽¹⁶⁾. Between AI and “collaborative creativity” ⁽¹⁷⁾ of the village ⁽¹⁶⁾ excellent processes can be developed. I believe that with the help of artificial intelligence it can and will become possible, if it is not already being tried. Having knowledge that all liquid processes could minimize solvent use would be giant leap to “Net Zero” processes. Developers can take advantage of point of addition and heat of reaction to control the reaction kinetics to speed up the processes. 

 

Having an all liquid process could lead to use of smaller size reactors or alternate reaction systems ⁽¹⁶⁾ that could eliminate or significantly reduce the amount of solvent use. It would be a significant step to achieving “Net Zero”. 

 

I believe that AI could scour the processing equipment world and propose equipment that is commercially used in chemical and other industries and would be suitable for the chemical synthesis and blended product applications. It will be interesting to see how AI will change chemical engineering practices. My conjecture is that it could propose processing methods and sequences that would simplify chemical process manufacturing and further augment “collaborative creativity” ⁽¹⁶⁾ of the “village” ⁽¹⁵⁾. Business models should improve.  

 

For the AI proposed designs my expectation is that the unit processes and unit operations might not change much from traditional ways but not having seen any such design it would enlightening to see the results of what will change. Proposed process designs would be readily accepted in the companies after they have been tested internally and product quality meets or exceeds expectations. Adoption in most companies will come quickly. However, companies that are regulated e.g. pharmaceuticals will be skeptical and apprehensive. This will be due major regulatory acceptance challenges as most of the regulatory folks have essentially no or very little chemical process development, design, commercialization and manufacturing experience. Lack of the knowledge and experience at regulators could be a major expenditures and time delays as the regulators will have to be taught value of better technologies. There will be cost implications to the companies and they might not or be slow to venture out in using AI based designs for their product development and manufacturing.  

 

Brand pharma companies could adopt the AI based technologies quickly as they are single product focused. However, generics since they are fragmented and do not have dedicated equipment could be reluctant to accept better technologies that could be AI generated. If AI based manufacturing is accepted overall equipment running time in pharmaceuticals could be lower than the current times. This might deter adoption of AI based technologies. Regulations would have to re-written. 

 

With the progress that is being made in AI it is expected that process development and manufacturing technologies will be of great benefit to humanity ^{(18, 19)}. With the changing landscape regulators will have change their posturing especially when it comes to pharmaceutical product approvals. If they don’t and will not keep up with time, stone age will be a burden on space age needs. 

 

Tomorrow is here:

 

Some might say AI will arrive tomorrow. However, we all know that TOMORROW never comes. AI’s TOMORROW is HERE. We have to learn and capitalize on what we do especially when it comes to chemical products and processes. Listening to the linked ^{(20, 21, 22, 23)} my conjecture is that it will have a powerful impact on chemical process synthesis and could deliver us pathways and processes that we think are or thought to be impossible. Acceptance in chemical process development, design and commercialization could be fast in non-FDA approved products. Incorporation of AI in drug development could be fast. However, as suggested earlier, incorporation of AI in manufacturing could lag “quite” some time due to regulator’s lack of hands on knowledge and experience in product, process development, design and manufacturing practices. My conjecture is that combination of “Creative Destruction” ^{(24, 25)} and “Nondestructive Creation” ^{(16, 26)} will come in play. Business Models, Regulatory policies and procedures will require a monumental change. 

 

Girish Malhotra, PE

 

EPCOT International

 

1.     Artificial Intelligence, Wikipedia  https://en.wikipedia.org/wiki/Artificial_intelligence Accessed March 21, 2023

2.     Chemical Engineering And Artificial Intelligence https://aichatgpt.co.za/chemical-engineering-and-artificial-intelligence/ Accessed March 21, 2023

3.     Artificial Intelligence In Chemical https://aichatgpt.co.za/artificial-intelligence-in-chemical-engineering/#AI_For_Developing_Novel_Chemicals_and_Products Accessed March 22, 2023

4.     Stephanopoulos, G. Artificial Intelligence in Process Engineering, Chemical Engineering Education https://journals.flvc.org/cee/article/view/124525/123536 pg. 182-185, 192 Accessed March 21, 2023

5.     Lou, H. H., Gai, H. Lamar University, How AI can better serve the chemical process industry, July 2020 Accessed March 21, 2023

6.     Trinh, C.; Meimaroglou, D.; Hoppe, S. Machine Learning in Chemical Product Engineering: The State of the Art and a Guide for Newcomers. Processes 2021, 9, 1456. https://doi.org/10.3390/pr9081456 Accessed March 21, 2023

7.     Gasteiger J.: Chemistry in Times of Artificial Intelligence ChemPhysChem 2020, 21, 2233– 2242 Accessed March 21, 2023

8.     Burke, J. What does net zero mean? https://www.greenbiz.com/article/what-does-net-zero-mean, May 2, 2019 Accessed April 27, 2021

9.     Sheldon R.A. The E factor 25 years on: the rise of green chemistry and sustainability, Green Chemistry https://pubs.rsc.org/en/content/articlelanding/2017/gc/c6gc02157c/unauth#!divAbstract , 2017, 19, 18-43 Accessed February 17, 2021

10.  Chemical Abstract Service https://www.cas.org/cas-data Accessed March 21, 2023

11.  NIST Chemistry Web Book SRD 69 https://webbook.nist.gov/chemistry/ Accessed March 21, 2023

12.  Malhotra, Girish: Process Simplification and The Art of Exploiting Physical Properties, Profitability through Simplicity March 10, 2017 Accessed March 21, 2023

13.  Malhotra, Girish: Sociochemicology, 2014 Accessed March 21, 2023

14.  AI for chemistry https://chemintelligence.com/ai-for-chemistry Accessed March 21, 2023

15.  Butler, Keith T. et.al Machine learning for molecular and materials science, Nature volume 559, pages 547–555 (2018) Accessed March 22, 2023

16.  Malhotra, Girish: Active Pharmaceutical Ingredient Manufacturing: Nondestructive Creation April 19, 2022 https://www.degruyter.com/document/doi/10.1515/9783110702842/html

17.  Malhotra, Girish: API Manufacturing and Environmental Sustainability Chemistry Today, September/October 2022 Vol. 40(5)

18.  Artificial Intelligence In Chemical Engineering https://aichatgpt.co.za/artificial-intelligence-in-chemical engineering/#AI_For_Developing_Novel_Chemicals_and_Products Accessed March 22, 2023

19.  Chemical Engineering And Artificial Intelligence https://aichatgpt.co.za/chemical-engineering-and-artificial-intelligence/ March 21, 2023

20.  A360 Day 1 Emad Mostaque https://www.youtube.com/watch?v=aEWHrqxniLM Accessed March 24, 2023

21.  Bubeck S. et.al. Sparks of Artificial General Intelligence: Early experiments with GPT-4, https://arxiv.org/pdf/2303.12712.pdf Accessed March 24, 2023

22.  Gates Bill: AI and the rapidly evolving future of computing https://www.youtube.com/watch?v=bHb_eG46v2cAccessed March 24, 2023

23.  Gates Bill, The Age of AI has begun https://www.gatesnotes.com/The-Age-of-AI-Has-Begun March 21, 2023 Accessed March 25, 2023 

24.  Malhotra, Girish: Is "Creative Destruction" the way to go for the Pharmaceuticals? Profitability through Simplicity, December 12, 2008 Accessed March 22, 2023

25.  Creative Destruction Schumpeter: Definition https://youmatter.world/en/definition/creativedestruction-schumpeter-definition/ April 20, 2020 Accessed August 15, 2022 

26. Kim, Chan W., Mauborgne: Nondisruptive Creation: Rethinking Innovation and Growth, MIT Sloan Review, February 21, 2019

Secret Life of APIs

Included in this post are some articles I posted on Pharmaevolution, an old Informa company. This site is defunct so I decided to re-post them on my blog. Some of the references are not available as the posting companies have changed hands.  
 
Tuesday, February 26, 2013 

The Secret Life of APIs

Active pharmaceutical ingredients (APIs) make the pharma world go 'round. Without them, any finished drug would be a mere placebo.

So why is it that 99 percent of the discussion taking place about pharmaceutical quality by design (QbD), process analytical technology (PAT), and other “new” methodologies (actually, there’s nothing new about them, and those of us in the know have been using these methods for at least four decades, but I digress) focuses on finished drug products, and leaves APIs out entirely?

In this blog series, I’ll try to right this wrong.

My goal is to share my views on the API value chain and the issues affecting their development, manufacture, and distribution. How might we develop and manufacture APIs at minimal cost and optimal quality, the first time and all the time?

In other words, I’ll explore how we can bring a QbD approach to the API world.

First, we have to accept the fact that QbD is not a technology toolkit (I’ve been horrified by the number of articles I see describing QbD as a technology platform!) but, rather, a disciplined way of incorporating the fundamentals of chemistry and chemical engineering. Done right, it will create a process that yields quality product consistently, and despite variations in raw material sources, but will meet established quality specifications.

Through these posts, I’ll present my view of the ideal path to consistently safe and high-quality APIs. Some readers will disagree with my views, and some will have better ideas. I hope that you will share them! My intent is not to challenge, but to review a process that will be required if we are to continue to produce quality products, consistently, at the lowest cost, using sustainable and safe processes.

Surely our creativity, imagination, and ingenuity will allow this to happen. But this process can’t just start at the end of the chain. It must start with the API, which, after all, defines each drug.

First, let’s quickly review API manufacturing. Active content in a tablet ranging from fractions of a milligram to hundreds of milligrams results in variable annual demand for APIs (from a few to more than 100,000 kilograms). In isolation, these numbers might not mean much, but they are critical in process design and production planning.

Unfortunately, other things get in the way, too, particularly the reliance on “Quality by Analysis,” or the common practice of analyzing every step of a reaction process to ensure that the process is progressing as expected. This “Analysis Paralysis” approach is already inefficient. But it stifles innovation and creates ripples of inefficiency in other business areas, including how we manage our supply chains and use capital equipment and other assets.

Under “AP” mode, our focus is on complying with regulations for safety, health, and the environment, rather than optimizing our processes so that they exceed regulatory expectations and enable continuous improvement.

As engineers, chemists, and other trained scientists, we learn the fundamentals that allow us to use QbD methods to design processes. However, we still don’t incorporate them fully into day-to-day operations. This isn’t because they’re not good ideas, but because business practices blind us to the need to optimize.

At a time when blockbusters (drugs whose annual sales are a billion or more dollars) are coming off patent right and left, why are we are still stuck in the mindset of finding and developing new blockbusters? Some new drugs, recently approved by FDA, serve only 3,000 patients worldwide.

In the early 20th century, demand was focused on the developed nations. Patents could be fully enforced, and there was minimal threat from developing nations.

Since the single API volume per site was generally low, it made sense to manufacture the APIs in batch mode, using the same or similar equipment that was used to produce non-drug products and fine or specialty chemicals. There was minimal consideration about the manufacturing technologies, because the costs related to inefficiency and regulatory compliance could be passed on to the customers. Profits were high, as the pricing was based on need rather than competition.

In the early 2000s, the playing field began to change. With patents expiring or being challenged, more countries placing price controls on pharmaceuticals, and new drug pipelines drying out, many companies are scrambling to prop up their balance sheets. One way they do this is through acquisitions, an unproven strategy that often yields only short-term results. Acquisitions are not going to increase pharma's new drug development success rate.

Business realities are changing rapidly today, and pharma must capitalize on opportunities that it has been ignoring for decades. The time has come to bring innovation to both API and finished drug manufacturing, and to bring more jobs back onshore, whether that shore is North America, Europe, or (as costs of doing business there continue to increase) China or India.

Are you with me? We will follow this with a brief questionnaire to help us discuss and shed more light on current practices in the industry. Look for it next week!

Girish Malhotra, PE 

EPCOT International 

•    Neglected Tropical Disease Drugs: What Are They Telling Us About Innovations?
•    Opportunities vs. Reality

April 4, 2013

The Secret Life of APIs: The Importance of Sociochemicology

When we produce a small-molecule active pharmaceutical ingredient (API), we have to pay close attention to its manufacturing process and quality systems, giving these much more scrutiny than we would for other synthesized chemicals. After all, APIs have the potential to improve our health or extend our life, where chemicals merely facilitate our lifestyles.

Thus, for API manufacture, we must have complete command of what we do. First-time quality has to be the way of life and our goal. Anything less than a quality-by-design (QbD) approach will add cost to the product.
Each chemical involved in the API manufacturing process is either a reactant, solvent, intermediate, or byproduct, and each has its own physical and chemical properties (for example, physical state, molecular weight, density, phase transition temperatures, solubility in water and other chemicals, viscosity, surface tension, and heat of formation). Some APIs are converted to a salt for easier dissolution, improved efficacy, and performance.

Each also has its own toxicity and toxic behavior when interacting with other chemicals. Since chemicals, like animals, have individual and collective behavior I call this behavior “sociochemicology.”
I'm not trying to oversimplify what all chemists and chemical engineers learn. However, we often seem to fail to account for the individual and collective behavior of constituents in our APIs. QbD becomes second nature when we understand and manage the behavior of chemicals that are used in the manufacture of active ingredients and formulations.

With this knowledge, we can manage, juggle, manipulate, maneuver, and even coerce behavior to develop a process that will produce quality product from the onset -- i.e., achieve the goals of QbD.

Such processes will be simple. We will have command of each reaction step and, collectively, for the whole process. In addition, the resulting process will offer the highest product yield.

Chemists and chemical engineers use sociochemical properties to select the desired chemicals for specific reactions. This information also allows us to have a safe starting point for reactions. We can use our knowledge and imagination to alter their behavior to our advantage. Understanding collective behavior of chemicals is important, because it allows us to handle them safely and properly in a process. It also gives us clues about how to manipulate and modify their relative amounts and reaction conditions to achieve the highest product yields, and the most economical and sustainable processes.

We can use sociochemicology to optimize process chemistry and operating conditions. Design of experiments should be used for optimization.

Reactivity and behavior can also be used to simplify addition point and method. Mutual solubility or lack of it, for instance, can facilitate phase separation or the removal of a desired product from the reaction mass. Solubilities also assist us in minimizing the number of solvents used. Process conditions can be used to influence reaction rate, flow, and solvent amount. Productivity, investment, and product cost are influenced. Physical state guides us to select the best flow control method. Clear liquid is the easiest to control, with gas being the hardest. Exotherm is best controlled using a heat exchanger vs. adding water/solvent or ice to the reaction mass.

Understanding of sociochemicology helps developers exploit unit operations to create economical, sustainable, and safe processes.

If we follow these basic principles, we have an opportunity to exceed regulatory guidelines in every step of API synthesis.

Additional reading:

·       Malhotra, Girish: Chemical Process Simplification: Improving Productivity and Sustainability, ISBN: 978-0-470-48754-9, January 2011, John Wiley & Sons Inc.
·       Malhotra, Girish: Focus on Physical Properties to Improve Processes: Chemical Engineering, Vol. 119 No. 4 April 2012, pgs. 63-66
 
Girish Malhotra, PE

EPCOT International 

April 9, 2013

Sociochemicology


Girish Malhotra explains "Sociochemicology," how to exploit chemical properties to optimize API development, and the potential savings involved, all in under three minutes!
Malhotra, Girish: Sociochemicology May 30, 2013 Accessed January 13, 2023

Girish Malhotra, PE

EPCOT International