Applying mathematical concepts to biological systems can speed up the drug development cycle through a better understanding of cell metabolism.

Prashant Pokhriyal, our next pathbreaker, works as Data Scientist at Yokogawa Insilico Biotechnology GmbH, a technology-based company that provides products and services for its enterprise digital-twin technology focused on biopharmaceuticals.

Prashant talks to Shyam Krishnamurthy from The Interview Portal about his first encounter with Bioprocessing technologies that prompted him to change his engineering branch from Electronics & Communication to Biochemical Engineering.

For students, Machine learning algorithms are all set to change the way drugs are developed and manufactured, but as they say, with opportunities come challenges that need to be solved !

Prashant, Your background?

I was born and brought up in Corbett City Ramnagar, Uttarakhand. My father is a retired Indian Army personnel, while my mother is a homemaker. I also have an elder sister who has been educated in agricultural sciences and management. I was a sincere student throughout my school. Rather than scoring high marks, I was more interested in unraveling mysteries of the universe, and so studying science came instinctively to me. I selected both mathematics and biology, along with physics and chemistry as my subjects during my senior secondary, which created the foundation for my future career in biotechnology. Apart from studies, I was a part of the basketball team and represented the school in the tournaments as a player. 

What did you do for graduation/post graduation?

I completed my Bachelor’s in Biochemical Engineering in 2016 from BTKIT Dwarahat, an autonomous engineering college under Uttarakhand Technical University. I then took admission in the prestigious Institute of Chemical Technology, Mumbai and graduated with a degree in Bioprocess Technology in 2018.

What were the influences that led you to such an offbeat, unconventional and unusual career?

Initially I had opted for a BTech in Electronics and Communications Engineering stream out of peer pressure, & because of better future opportunities in IT and Electronics in India. During my first year as an undergraduate, an encounter with the book ‘Bioprocess Engineering’ by Michael Shuler describing the origin and perspective of Bioprocess engineering enamored me. I was excited to read about the fascinating world of Bioprocessing. I was fascinated, especially by the history behind the large-scale production of Penicillin, which played a huge role during World War 2. Even though everyone around me didn’t consider Biotechnology in high esteem because it was more research-focused and had a bleak job scenario, in my second year of my undergraduate studies, I changed my engineering branch to Biochemical Engineering. In hindsight, I have never regretted my decision. 

I got admission to one of the country’s best colleges for my Master’s and worked under one of the experts. 

How did you plan the steps to get into the career you wanted? Or how did you make a transition to a new career? Tell us about your career path

I was able to secure a decent score in the GATE exam and get admission into the MTech program in Bioprocess Technology at Institute of Chemical Technology, Mumbai, a college focused on imparting research education in chemical engineering and allied disciplines. ICT has churned out several industrialists in the field of chemical engineering like Mr. Mukesh Ambani (Reliance Industries) and Dr. KH Gharda  (Gharda Chemicals) throughout its illustrious history. I liked the industry-academia collaboration in my institute, and the technologies that the institute produced which were implemented by industries. 

During my Master’s project, I worked on both metabolic and kinetic models to obtain industrially relevant products. In that project, I applied Flux Balance Analysis (FBA) on E-Coli to optimize overproduction of acetate, using the COBRA toolbox in the MATLAB environment, and additionally built a kinetic growth model for microalgae. The projects involved performing wet-lab experiments as well as mathematical optimization and basic computer programming. 

How did you get your first break?

After graduation, I joined the industry. I find it fulfilling when I realize that my work directly makes a real impact in the world. 

Initially, I executed basic cell culture techniques: Inoculum preparation, passaging, operation of shake flasks and Bioreactors etc. Thereafter, I performed extensive experimentations in HPLC and UPLC based analytical techniques. Additionally, I have also worked on the preparative techniques for purification of antibodies like TFF, Process scale chromatography, refolding etc. 

After that period, I was assigned to evaluate new technologies and to create value out of existing workflows through the application of statistical, machine learning and/or mathematical modeling. 

Biosimilar drugs, which are the officially approved versions of innovator drugs, because of their short-development time and being up to 50 % cheaper, are a boon especially for low-income & developing countries. The key for their development is faster development timeline and regulatory approval. I was involved in the different departments of R&D, ranging from protein production, purification and analytical sciences. 

What were some of the challenges you faced? How did you address them?

Challenge 1: Machine learning algorithms require lots of good quality data in the correct format. However, the data infrastructure in biopharma industries still leaves a lot to be desired. All this leads to arduous pre-processing steps.

Challenge 2: Applying mathematical concepts to biological systems, and balancing the two is the next big challenge. Biopharmaceutical experimental data is generated from analytical instruments having measurement uncertainty, which adds uncertainty on top of uncertainty generated by machine learning algorithms. Oftentimes, we must deal with incomplete data, and we are required to make smart assumptions going forward building digital twins.

Challenge 3: Since I work at the intersection of biological systems and machine learning, I am expected to be proficient in and be up-to-date with new research undertaken in both fields. Oftentimes during work, I stumble upon some new concept in either domain, which I must learn before proceeding forward.

Where do you work now? What problems do you solve?

I work at Yokogawa Insilico Biotechnology GmbH which is a technology-based company that provides products and services for its enterprise digital-twin technology focused on biopharmaceuticals. The features of the technology enable our customers to optimize their process development through feed & media design, process parameters etc., increase process knowledge and scale-up from lab to manufacturing. The digital twins follow a hybrid modeling approach and consist of a mechanistic metabolic network model (metabolic reactions of the cellular network), process model (mass-balance equations) coupled with a data-driven model (machine learning algorithms). 

I am a part of an application scientists’ team which builds digital twins for Upstream process development for Biopharma organizations. We are the client-facing team of the organization and take raw data from our clients and feed them into our digital twin workflows, and get a digital twin of that process. Now the digital twin can be used to predict the behavior of the cell metabolism and process conditions. In most cases, we provide our customers with the optimal process conditions (amino acid composition, temperature etc.) for maximization of titer (concentration of molecule of protein of interest) as objective function.

What are the skills required for your role?

For this job role, one should have a thorough understanding of cell metabolism and it is good to have experience in laboratory techniques in cell culture. One necessary skill is to have hands-on experience in the Pandas library of Python, which is used for the data pre-processing. A good grasp on statistical and machine learning concepts is also a must-have skill. Additionally, good communication and presentation skills would help in better alignment with the customer.

What is a typical day like?

A typical day varies according to the stage of the project. Some days are spent entirely on programming our data pipelines. We sometimes also spend time with sales, marketing, and business development colleagues to discuss commercial matters, and align with the product development & software teams related to our machine learning pipelines. 

I like working in a cutting-edge interdisciplinary field, with a large scope of learning, and interacting with a diverse team.

How does your work benefit society? 

A typical drug development program can cost up to billions of dollars. With our digital twin solutions, we not only improve productivity of the cells, but also provide mechanistic understanding of the cell metabolism, which is crucial during the scale-up and regulatory submission processes. Using our solutions, our clients can produce a drug with higher profit and with much less effort. Even a 6-month reduction in the drug development pipeline can increase the profit of biopharma organizations by millions.

Tell us an example of a specific memorable work you did that is very close to you!

Once we were trying to build a scaled-down model for a chromatographic unit operation for protein purification. We couldn’t find much literature on the topic, and I was also not much experienced with machine learning at the time. We still went ahead, and after several brainstorming sessions and learnings, met our objective. We were able to publish our research in a leading journal of our field. 

Your advice to students based on your experience?

With failures comes wisdom. One must follow his/her passion, but also be prescient enough to know what the future holds. Grind and focus are the keys of mastering any skill, much more than raw talent. 

Future Plans?

I plan to improve my skills especially in the field of machine learning. Additionally, I would like to explore commercial aspects of data analytics in the future.