Prediction of evolutionary constraint by genomic annotations improves prioritization of causal variants in maize

Authors: RAMSTEIN, GUILLAUME - Aarhus University; Buckler, Edward - Ed

Submitted to: bioRxiv
Publication Type: Pre-print Publication
Publication Acceptance Date: 9/5/2021
Publication Date: 9/5/2021
Citation: Ramstein, G.P., Buckler IV, E.S. 2021. Prediction of evolutionary constraint by genomic annotations improves prioritization of causal variants in maize. bioRxiv. 2021.09.03.458856. https://doi.org/10.1101/2021.09.03.458856.
DOI: https://doi.org/10.1101/2021.09.03.458856

Interpretive Summary: In plant breeding, associations between DNA variants and agronomic traits are useful to select the most promising varieties. However, these associations are only correlations and they cannot tell us what exact DNA mutations cause the observed differences in agronomic traits. In this project, we proposed a strategy to accurately detect the effect of DNA mutations. Instead of estimating statistical associations, we used evolutionary conservation across species to detect impactful mutations. In maize, we showed that this approach can detect effects of DNA mutations on fitness. Moreover, we showed that novel machine learning techniques can further improve the accuracy of our detections. Our method will allow plant breeders to target specific mutations for breeding applications like CRISPR-based editing, which require accurate detection of causal changes in the DNA. It can support this technology by guiding CRISPR-based editing against the most disadvantageous mutations in crop genomes. Therefore, our proposed strategy can accelerate genetic gains for important fitness-related traits, like grain yield or resilience, in species like maize or in understudied crops, which carry many disadvantageous mutations.

Technical Abstract: In plant breeding, associations between DNA variants and agronomic traits are useful to select the most promising varieties. However, these associations are only correlations and they cannot tell us what exact DNA mutations cause the observed differences in agronomic traits. In this project, we proposed a strategy to accurately detect the effect of DNA mutations. Instead of estimating statistical associations, we used evolutionary conservation across species to detect impactful mutations. In maize, we showed that this approach can detect effects of DNA mutations on fitness. Moreover, we showed that novel machine learning techniques can further improve the accuracy of our detections. Our method will allow plant breeders to target specific mutations for breeding applications like CRISPR-based editing, which require accurate detection of causal changes in the DNA. It can support this technology by guiding CRISPR-based editing against the most disadvantageous mutations in crop genomes. Therefore, our proposed strategy can accelerate genetic gains for important fitness-related traits, like grain yield or resilience, in species like maize or in understudied crops which carry many disadvantageous mutations.