Bayesian analysis improves legume varieties

Bayesian analysis is a systematic statistical means that can harness the data of previous on-farm trials to better understand how to make the genetic selection process more efficient in creating improved crop varieties.

While legumes are vital to meeting the nutritional needs of Afghanistan’s population, their uptake as an agricultural product and their harvest yields remain low under farmers’ conditions. As a result, ICARDA and Afghanistan’s Ministry of Agriculture, Irrigation, and Livestock (MAIL) collaborated on a series of demonstration on-farm trials in 2009-2012 throughout three provinces to showcase improved seed varieties and best agronomic practices. Using that data to refine the means by which improved varieties are created, the method provides an empirical and evidence-based path towards agricultural breeding.

Up until recently, the data from such trials would be largely underutilized, if not entirely ignored. Much like designing a new drug and not using any clinical trials data that came beforehand, such an approach neglects to build on previous successes and failures. Thus, because farmers’ preferences heavily influence what they plant and how they approach agro-technology methods — thus throwing more random variables into the process — the Bayesian approach seemed a more appropriate way to analyze the data.

The yields of the demonstrations were compared with the yields of farmers growing local varieties with local agronomic practices. The datasets for 2012 were evaluated to compare the seeding and agro-technologies employed for productivity and risk, while the data from 2009 to 2011 was used to establish variance parameters for the analysis.

In this case, Bayesian analysis was used to calculate the probability distribution of various parameters to refine and identify accessions' potentials and to predict the genetic gain from crossing particular genotypes to achieve desired traits — for example, high-yielding wheat with drought tolerance.

Using empirical evidence to inform the iterative process of breeding selection further validates the evidence-based grounding of crop breeding and builds on the breeding successes (or failures) to improve uptake and productivity.