Paul Skroch VP Data Science and Platform Engineering, Benson Hill

Paul joined Benson Hill in 2016 as Vice President of Data Science and Platform Engineering overseeing the data science and systems biology groups as well as software development.

At Benson Hill, we believe all breeders should be able to leverage the most cutting-edge advances in data science and plant biology to improve any crop, regardless of organization size. A modern food system needs the power of prediction to be in the hands of breeders, everywhere.

The Benefits of Machine-learning, Data-driven Prediction

Tremendous progress has been made in outcome prediction across many industries thanks to an explosion of data, advances in machine learning science, and improvement in cloud computing and hardware acceleration. From healthcare to retail to entertainment, prediction is allowing innovators to better develop and deliver products. In precision farming, prediction can define optimal agronomic practices, but there are more fundamental ways that prediction can improve crops.

Prediction can optimize crop attributes, creating new and differentiated food and ingredients that consumers demand as part of a healthy, sustainable food system. Continued genomics advances have significantly driven down the cost and time required to generate high quality information. Paired with advances in phenotyping and environmental monitoring, this data explosion has opened new opportunities in agriculture. 

Prediction will Unlock Product Innovation

Predictive breeding is one of the most powerful ways that innovators can capitalize on the massive influx of data to develop new crops and capture new market opportunities. While prediction has long been used in plant breeding performance trials, the process has been costly and time-consuming. Over the last decade, traditional plant breeding practices have been augmented by genomic-based approaches that link molecular markers (genotypes) to trait variation (phenotypes), or even use whole genome estimates to improve the assessment of future genetic potential. 

This methodology, called genomic selection, is a form of predictive breeding that was first applied in the animal industry and is considered perhaps the most important technology since artificial insemination. Its implementation in a national breeding program of Holstein cows resulted in genetic gain increases ranging from 50-400% in just 7 years. Similar approaches are being successfully applied to plant breeding. 

As the richness and volume of data grows, and as data sets can be connected across locations and time, the true power of predictive analytics through machine learning can be unlocked to better connect decisions and predict outcomes. Until now, the growing potential of predictive breeding has not been accessible to many breeding programs.

Platform Technologies Empower Breeders

Benson Hill’s crop design platform, CropOS™, integrates many types of data including the genotypic and phenotypic data used in plant breeding and makes that data accessible for use in predictive analytics. These integrated data sets are used to train models and predict outcomes leveraging machine learning, simulation, and other statistical techniques. Key steps in the workflow are enabled through a highly usable web interface that allows breeders to extract more information from their data. More informed choices can be made without the need to learn data science and computer programming.  

In the past only a few large companies could afford the resources required to build a platform with this capability. Now, Benson Hill’s Breed application puts the power of genomic prediction at the fingertips of breeding programs of all shapes and sizes. Join us for a live demo of Breed on May 8th.