Location: Physiology and Pathology of Tree Fruits Research
Project Number: 2094-43000-008-017-S
Project Type: Non-Assistance Cooperative Agreement
Start Date: Sep 2, 2021
End Date: Sep 1, 2026
Objective:
Postharvest losses of pear and apple fruit quality are costly to the tree fruit industry. Despite layers of sophisticated postharvest technology, there are no current reliable risk assessment tools for losses of fruit quality during the postharvest period. Furthermore, production practices, the orchard environment, postharvest practices, and cultivar differences all play a coordinated role to specify fruit quality in the postharvest period - how these factors all interact is poorly understood.
Our goal is to find genetic factors related to postharvest fruit quality traits. This involves experimentally isolating genetic factors that play a role in important postharvest fruit quality traits. Currently, we are focused on searching for genetic factors that are related to the risk for postharvest losses in quality. Identification of these factors will allow us to better understand the molecular mechanisms that influence postharvest quality, giving us insight that could lead to enhanced management practices. This includes more precise application of existing technology, and the development of new technology to enhance postharvest fruit quality.
Our past work has shown that gene activity signatures are relatable to aspects of fruit physiology in the postharvest period. A significant hurdle to discovering key important genetic factors is the complex nature of postharvest fruit quality traits - large numbers of genes with complex patterns of expression and co-expression play a role. Therefore, this project aims to develop new methods to identify key genetic factors by analyzing massive gene activity datasets that influence or specify important postharvest fruit quality traits.
Objectives:
1. Build a comparative genomic framework to classify apple and pear genes into gene families.
2. Incorporate gene family information into gene discovery workflows to identify key genetic factors for fruit quality traits.
3. Validate the activity of selected genes across our catalog of fruit samples (multiple years/orchards/cultivars - in hand, and growing annually).
4. Current project objectives require rich fruit physiological information so as to link fruit quality characteristics to genes of interest. This amendment will support work to improve the quality of the fruit physiological data.
Approach:
Generally, the cooperator will combine horticulture methods and genomics methods to search for gene of interest. This includes the generation of fruit physiological data from field experiments and samples from industry cooperators. The Cooperator will analyze and interpret transcriptome and genome data using comparative genomics, phylogenomics, and various computational techniques, such as supervised machine learning, to identify genes and patterns of gene activity that are relatable to at harvest and postharvest fruit physiological data. The cooperator will then examine gene activity of candidate genes in additional fruit samples (from our large catalog of validation samples) to validate gene activity patterns for the genes of interest in a broader context (i.e., other years, cultivars, orchards).
Our current AI modeling approaches to identify gene that influence postharvest fruit quality tratis (a.k.a. gene activity-based biomarkers) include: random forest, gradient boosted trees and other exploratory methods. A software we are developing called Granny can produce visually-scored fruit physiological traits (e.g. fruit starch content) using an AI approach coupled with additional image processing. Integration of our current gene-trait models into Granny will consolidate our efforts into a single software package. This will enhance interactions with our partner growers. Additionally, it will allow partner growers to provide to us their in-house visual trait data. This in-house data is needed to adjust our models when starch ratings (human or AI predicted) are incorrect for a given year. Therefore, we request funds to partially support a postdoc to integrate AI-based starch ratings into Granny, and work with other project personnal for mobile app integration.
