By Matt Fates, Partner, and Jeff Knox, Associate
Summary
The world has grand ambitions to build much of tomorrow’s materials with biology. Forget just the fermentation-derived materials of today – the beverages, the food additives, the fuel ethanol, the cosmetics ingredients, the pharmaceuticals. Tomorrow, we want to lab-grow meat, and bio-plastic, and more. There are a lot of good reasons to build with biology, but to move toward this future we’ll need to increase our collective biomanufacturing capacity, and also better use the capacity we have.
We’re excited about BioIntelligence because they create that potential, to better utilize our biomanufacturing capacity. Precision fermentation, one important type of biomanufacturing, is a complex process that varies widely in its efficiency. There’s a lot of value in understanding if the process is going according to plan. However, many of our best tools for understanding the inner workings of a bioreactor aren’t actionable because they don’t work in real-time. We’re excited about BioIntelligence because their technology can produce data on biomanufacturing processes that is both accurate and real-time, enabling the optimal biomanufacturing methods that we’ll build with tomorrow. Founder Joel Sirois and his team have developed the interlocking hardware and software components of their solution leveraging years of academic research. They’ve impressed us with their early success in working with customers across multiple industries. We’re excited to back them
as they continue the work of putting AI to work for good in biomanufacturing.
The Problem
Already today biomanufacturing is used to build a greater variety of products than many may realize. Fermentation alone is used to produce material for a range of products and industries as diverse as food, flavors and fragrances, agriculture, industrials, and pharmaceuticals. Fermentation is used in so many places because it is versatile, cost-effective, and scalable. From relatively inexpensive feedstock you can produce a huge variety of products – just change the organism doing your fermentation.
However, for its versatility, fermentation is also time-consuming and inconsistent. Cell growth can create a bottleneck to production speed, and the process often gives significant batch-to-batch variation in yield based on very small changes to initial conditions. Together, these two drawbacks create painful inconsistency for manufacturers. It’s possible to run a 72- or 96-hour fermentation only to discover that your batch produced half the amount of product it was supposed to. These ‘deviations’ in manufacturing process are costly. Ideally a manufacturer would notice and correct them as they occur. To do that, though, the operator of the bioreactor would need accurate, real-time information to act on. Therein, for some time, lay the problem.
Advances in laboratory technology over the past decades have given us many good tools for analyzing various components of a fermentation broth – mass spectrometry, flow cytometry, and more are excellent tools. However, by and large, these tools require 1) a sample be taken from the fermentation vessel, and 2) time to run. The result is a clear picture of what the fermentation broth looked like minutes or hours earlier. That time delay makes such ‘off-line’ data much less useful to an operator trying to steer the progress of the fermentation, called feed-forward control.
The BioIntelligence Solution
We’re excited about BioIntelligence because they provide the real-time, high accuracy data that allows effective feed-forward control. BioIntelligence’s solution is a proprietary optical probe paired with bespoke AI analytical software.
BioIntelligence measures the contents of a bioreactor using light, harmlessly bouncing it off the fermentation broth and collecting it back at ever-so-slightly different wavelengths based on what’s inside. Biointelligence primarily tracks fluorescence, an especially bright type of returned light. BioIntelligence’s probe works even in cloudy reactor conditions and prevents clogging of the light-collecting slit of the optical sensor, two common problems.
Upon collecting the reflected light, BioIntelligence runs the data through an AI algorithm. Fluorescence signals are bright, but also noisy and not produced by every chemical. The algorithm distills the noisy signal to create highly accurate estimates of different features of the fermentation broth – the level of fermentable sugar, the amount of product generated, and so on. The BioIntelligence algorithm is effective out of the box, but can be improved further by tuning it to a specific fermentation vessel with 5-6 batches worth of offline data.
Most importantly, the system makes these estimates instantaneously. Accurate, real-time data allows an operator to understand what’s happening inside their fermentation vessel and react accordingly. By allowing operator to correct and optimize their processes, we can more efficiently build the bioproducts of both today and tomorrow.
Into the Future of AI-enabled Biomanufacturing
By enabling advanced process control of high-variation fermentation processes, BioIntelligence is poised to improve biomanufacturing in several domains. Generating better yields on our existing infrastructure will mean not just better margins for manufacturers, but better resource efficiency as well – averting waste of water and power. Joel and his team have developed their interlocking hardware and software technologies through multi years-long research and development processes, and we believe now is the time to put it into the hands of the manufacturers who can use it. We’re excited to support them as they put AI to work for the good of tomorrow’s bioeconomy.
Comments