Exploiting historical agronomic data to develop genomic prediction strategies for clonal selection early in Louisiana sugarcane variety development program

Authors: SHAHI, D - Louisiana State University Agcenter; Todd, James; GRAVOIS, K - Louisiana State University Agcenter; Hale, Anna; BLANCHARD, B - Louisiana State University Agcenter; COLLINS, K - Louisiana State University Agcenter; PONTIF, M - Louisiana State University Agcenter; BAISAKH, N - Louisiana State University Agcenter

Submitted to: The Plant Genome
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/18/2024
Publication Date: N/A
Citation: N/A

Interpretive Summary: Scientists are using a method called “genomic selection” to make sugarcane, grow better and produce more sugar. They looked at six different characteristics of the sugarcane, like how much sugar it can produce, how many stalks it has, and how heavy each stalk is. They used a lot of genetic information (20,451 SNPs) from 567 different types of sugarcane to make predictions about these characteristics. They found that they could predict some characteristics better than others. For example, they could predict how much sugar the sugarcane can produce and how heavy each stalk is well. But it was harder less accurate to predict how many stalks each plant would have. They also found that some methods, like “random forest” and “support vector machine”, were better at making predictions than others, especially for characteristics that are hard to predict. They also found that when they considered the traits taken in the early part of the program such as how many stalks a plant has and how much sugar it can produce together in a multi trait model, they could make better predictions about how much sugar a field of sugarcane can produce in the later stages of the breeding program. Finally, they found that they didn’t need all 20,451SNPs to make good predictions and that 9,091 were sufficient for the prediction of all traits. This study shows that using all this information and methods can help us choose the best sugarcane plants early on in the breeding program. This can help us grow better sugarcane varieties in potentially less time and produce more sugar.

Technical Abstract: Genomic selection can enhance the rate of genetic gain of cane and sugar yield in sugarcane, (Saccharum L.), an important industrial crop worldwide. We assessed the prediction accuracy (PA) of six traits, such as theoretical recoverable sugar (TRS), number of stalks (NS), stalk weight (SW), cane yield (CY), sugar yield (SY), and fiber content (F) using 20,451 SNPs with 22 different statistical models based on the genomic estimated breeding values (GEBVs) of 567 genotypes within and across five stages of Louisiana breeding program. TRS and SW, with high heritability, showed higher PA of sugar yield compared to other traits, where NS had the lowest PA. Machine learning (ML) methods, such as random forest (RF) and support vector machine (SVM) outperformed other models for predicting traits with low heritability. The PAs were higher in cross-stage validation than within stage five-fold validation. ML methods predicted TRS and SY with the highest accuracy in cross-stage prediction while Bayesian models predicted NS and CY with the highest accuracy. Extended GBLUP models accounting for dominance and epistasis effects showed slight improvement in PA for a few traits. When both NS and TRS, which can be available as early as the stage 2, were considered in a multi-trait selection model, the PA of SY in stage 5 could increase up to 0.66 compared to 0.30 with a single trait model. Marker density assessment suggested 9,091 SNPs were sufficient for optimal prediction of all traits. The study demonstrated the potential of using historical data to devise genomic prediction strategies for clonal selection in early stages of the Louisiana sugarcane breeding program.