Modelling GxE for the next generation of plant breeding

Modelling GxE for the next generation of plant breeding

We want to give a shout out to HIIT, the Helsinki institute for Information Technology and Jussi Gillberg et al. for their article titled “Modelling GxE with historical weather information improves genomic prediction in new environments”.

The more we work with our clients, the more data we see being gathered. Data in the field, data from the soil, environmental data, surveillance data … data, data, data.

Some of our clients are drowning in too much data and too little actionable insight of what to do “next”.  

As Gillberg and colleagues rightfully point out, the modeling GxE interactions can help answer a few of these “next” questions. Specifically: what to do when expanding to new markets, how do we make use of historical data, and maybe most importantly, how will we deal with climate change?

We agree with them: predictions in new environments are challenging, but worth while. They highlight how uncovering GxE will mitigate the problems of conventional breeding: accounting for historical weather in the actual target population of environment can prevent overfitting to the conditions where only a few field trials are performed. GxE models are also needed in assessing the effects of climate change and selecting varieties favourable to the new conditions.

Their methods, of course, are unique to them and the specifics of their experiment include predictions for totally unseen environments and genotypes. Even with such limitations, they point out that these predictions outperform current industry standards for the same predictions. 

We greatly respect and admire them for pushing the boundary in GxE predictions and the more publications that highlight the potential that is to be achieved in data incorporation and integration, the more the field will advance. We strongly believe in this, too! Our offering can integrate different datasets from more than 16 sources including genotyping, environment and phenotypes. Computomics is successfully using machine learning on biological data since 2012 and we have validated results in the field since 2015.

When innovation is on its way, to the unprepared it appears impossible… and as soon as it is mainstream, it seems obvious. It is the moment between those two instances when companies have the chance to make huge competitive gains.

Here’s to changing the world- one prediction at a time!

Be quicker than your competitors and talk to us to learn how we can help you be one step ahead!

Go back