Towards identification of postharvest fruit quality transcriptomic markers in Malus domestica

Authors: HADISH, JOHN A - Washington State University; Hargarten, Heidi; ZHANG, HUITING - Washington State University; Honaas, Loren; FICKLIN, STEPHEN - Washington State University; MATTHEIS, JAMES - Retired ARS Employee

Submitted to: Postharvest Biology and Technology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 12/27/2023
Publication Date: 3/6/2024
Citation: Hadish, J., Hargarten, H.L., Zhang, H., Honaas, L.A., Ficklin, S.P., Mattheis, J.P. 2024. Towards identification of postharvest fruit quality transcriptomic markers in Malus domestica. Postharvest Biology and Technology. 19(3):e0297015. https://doi.org/10.1371/journal.pone.0297015.
DOI: https://doi.org/10.1371/journal.pone.0297015

Interpretive Summary: Postharvest management of apple fruits is important for reducing crop loss and ensuring that consumers receive high-quality produce at the grocery store. An important aspect of this management is making sure that batches of apples go to market prior to them going bad. Knowing if a batch of apples is going to go bad in either 1 month or 5 months will allow packing houses to make marketing decisions that result in higher profits and higher quality fruit making it to market. Unfortunately, making these predictions is not always an easy task. Differences in the life history of the apple--i.e. weather, orchard location, soil type, pest pressure, and how the fruit are handled--all have an impact on how long they will last in storage. Measuring and interpreting how these life history events impact storage is difficult. Here we report experiments that use machine learning methods to create models that can predict loss of firmness in ‘Gala’ apples. Previous work, including our own, has shown that the life history events that influence how long an apple will store also influence that apple's gene activity patterns. Ever-advancing genetic technology allows deeper dives into gene activity patterns that we can relate to the final apple outcome. We brought together machine learning and modern genomics techniques in a proven horticulture research framework towards prediction of postharvest fruit quality in apples. This work also provides a roadmap for future investigations in other crops, not just tree fruit.

Technical Abstract: Predicting phenotypic traits is important in high-value agricultural crops where knowledge of disease, physiological traits, and disorders can reduce crop loss and improve end-use quality for consumers. These phenotypic traits are caused by the interaction of a crop’s genotype with its environment. In apples, negative phenotypic outcomes are often not observable at the time of harvest, and instead develop after the fruit have been in storage. Predicting these negative outcomes prior to their development would allow producers to make marketing decisions to reduce crop loss and maximize end-use quality. One promising method for predicting phenotypic outcomes is through measurements of the transcriptome, as it responds to environmental factors before visible phenotypes occur. Prognostic Transcriptomic Biomarkers (PTBs) are transcriptomic measurements of specific genes used to predict a phenotypic trait. The following research conducts a preliminary analysis to explore methods and approaches for the development of PTBs in apples by using loss of firmness in ‘Gala’ Apple Fruits. We compare Random Forest and Elastic Net’s ability to select PTBs from a large time-series RNA-seq data set that includes multiple postharvest treatments.