Location: Plant Science Research
Title: Multi-trait modeling and machine learning discover new markers associated with stem traits in alfalfaAuthor
![]() |
MEDINA, CESAR - University Of Minnesota |
![]() |
Heuschele, Deborah - Jo |
![]() |
ZHAO, DONGYAN - Cornell University |
![]() |
LIN, MENG - Cornell University |
![]() |
BEIL, CRAIG - Cornell University |
![]() |
SHEEHAN, MOIRA - Cornell University |
![]() |
Xu, Zhanyou |
Submitted to: PLoS Genetics
Publication Type: Peer Reviewed Journal Publication Acceptance Date: 7/30/2024 Publication Date: N/A Citation: N/A Interpretive Summary: Improving alfalfa digestibility is essential for enhancing milk and beef production. However, the stems of alfalfa have low digestibility, which accounts for 50 to 70% of the plant's composition. To address this issue, researchers have proposed marker-assisted and genomic-based selection as a viable method for improving the genetic gain of complex traits, such as alfalfa stem digestibility. Through multi-trait modeling and two machine learning algorithms, this study has identified new markers associated with stem traits in alfalfa. These markers can be utilized for marker-assisted breeding, which can lead to increased alfalfa stem digestibility, ultimately improving milk and beef production. Technical Abstract: Alfalfa biomass can be fractionated into leaf and stem components. Leaves comprise a protein-rich and highly digestible portion of biomass for ruminant animals, while stems constitute a high fiber and less digestible fraction, representing 50 to 70% of the biomass. However, little attention has focused on stem-related traits, which are a key aspect in improving the nutritional value and intake potential of alfalfa. This study aimed to identify molecular markers associated with four morphological traits in a panel of five populations of alfalfa generated over two cycles of divergent selection based on 16-h and 96-h in vitro neutral detergent fiber digestibility in stems. Phenotypic traits of stem color, presence of stem pith cells, winter standability, and winter injury were modeled using univariate and multivariate spatial mixed linear models (MLM), and the predicted values were used as response variables in genome-wide association studies (GWAS). The alfalfa panel was genotyped using a 3K DArTag SNP markers for the evaluation of the genetic structure and GWAS. Principal component and population structure analyses revealed differentiations between populations selected for high- and low-digestibility. Thirteen molecular markers were significantly associated with stem traits using either univariate or multivariate MLM. Additionally, support vector machine (SVM) and random forest (RF) algorithms were implemented to determine marker importance scores for stem traits and validate the GWAS results. The top-ranked markers from SVM and RF aligned with GWAS findings for solid stem pith, winter standability, and winter injury. Additionally, SVM identified additional markers with high variable importance for solid stem pith and winter injury. Most molecular markers were located in coding regions. These markers can facilitate marker-assisted selection to expedite breeding programs to increase winter hardiness or stem palatability. |