The story of POS Bio-Sciences’s (POS) venture into machine learning began with asking a very simple question: why not use all of the data collected over the years to predict future actions? While this question sounds like a radical way of approaching ingredient science, it is exactly the type of thinking POS decided to implement to help guide innovation in producing new ingredients and new proteins.

In the past, in-depth research and development (R&D) took ingredient scientists and process engineers a lot of time and energy, and would often not yield successful results. Previously, scientists’ personal experience and unique knowledge on raw materials, functional products, and ingredients, combined with extensive research using current literature, were key to R&D. Scientists would use their collective knowledge and repeatedly test the production process, tweaking all the elements to finally come up with an optimal process to produce a suitable ingredient. This was a very research-intensive and cost-intensive process, but what if there was a more efficient program that could help streamline this process and guide scientists to a successful result just from knowing the raw materials involved or the desired health benefits?

This is where POS’s machine learning technology comes in. With extensive data for innovative ingredients and processing work with proteins, oils, and starches, POS decided to use the background information they had to help streamline their development of new ingredients. Computational Scientist Dr. Surajith Nalantha Wanasundara (left) and several others at POS have worked hard to construct a program that records and compiles information from the generic background processing data that POS has gathered and generated over its many years of R&D work. Harnessing this data, the program can predict what feed stock and processing parameters are needed to achieve a certain health or functional benefit from an ingredient. In its simplest form, the database searches its information and predicts what raw materials and processing methods are needed to optimize the taste, texture, appearance, and cost of an ingredient. Using the database, scientists will be able to simply search a specific health benefit, and all of the information on the raw materials, as well as the processing methods needed to achieve an ingredient with that specific health benefit, will be displayed. They can also further refine their search by including greater detail. For example, they can search for a specific health benefit that comes from plant protein to be an alternative to animal protein.

By harnessing big data to facilitate innovation, the system has the potential to save ingredient scientists and processing engineers hundreds of hours in R&D and discovery work. Scientists and engineers can now answer the question – “where can we find the raw material, and how can we process it?” – at the beginning of their search, with POS’s new software pointing them in the right direction early on. With innovations like these, POS can maintain its competitive advantage and stay on top of changing health trends, especially when it comes to proteins. This platform allows POS to create new opportunities by using innovative processing techniques for value-added protein ingredients and helps streamline POS’s commercialization of new ingredients.

With information constantly being added, the program will continue to learn about the best methods of processing for each different raw material and resulting ingredient. Like a brain, the program learns from the data to get better at its own job. “The interesting thing about this database is that it is never complete,” says Nalantha. “We will keep adding data and information to make the predictions more accurate. More data equals better answers.”

POS continues to work on this innovation and others like it so that their partners and clients can master R&D and streamline efforts when it comes to producing innovative ingredients.


View All Blogs