Non Technical Summary
Fresh fruits and vegetables are an important component of the human diet, but their quality often deteriorates before reaching consumers due to ongoing biochemical processes and compositional changes.Freshness is an important and fleeting state, and its progressive loss after harvest cannot be easily evaluated. Until now this race against time has been almost entirely subjective, mostly based on external visual criteria. To the consumer, freshness is synonymous with proximity to the harvest time and retention of nutritional value.The current lack of objective indices for definingfreshnessof fruits or vegetables limits our capacity to control product quality and leads to food waste. In the proposed work,we will combine machine learning technologies and multi-omics tools to understand post-harvest senescence and microbial spoilage of fresh produce and then develop a simple imaging approach (FreshID) to evaluate fruit and vegetable quality, particularly nutritional and texture-related attributes such as maturity, storage ability, and firmness.This proposed comprehensive research program will identify the genes, proteins and compounds that can serve asfreshness-indicators and will then use these indicators todevelop of an algorithm to accurately estimate the freshness and/or produce spoilage.The outcomes of this proposed research will provide mechanisms for early detection of poor-quality produce and, by leveraging these mechanisms throughout the food production pipeline, understand how food waste can be reduced. Additionally, the FreshID toolbox of molecular markers associated with plant senescence can be used duringbreedingfor pre-harvest traits that could delay postharvest senescence. The goals of the proposed research are advances in both basic research and applied plant science that will impactfood security and human health. The developed tool would allow a new level of post-harvest logistics, supporting availability of high-quality, nutritious, and fresh produce.
Animal Health Component
Research Effort Categories
Goals / Objectives
Our specific objectives are:1) Discover invisible post-harvest changes that impending senescence and microbial spoilage in vegetables and fruits using imaging analysis;2) Identify genes, protein markers and chemical compounds that indicate freshness;3) Correlatehyperspectral imaging data with freshness indicators to develop an approach, FreshID.The FreshID approach has the potential to define afreshness signaturethat occurs prior to visible deterioration.A long-term goal of the proposed research is to aid development of an innovative FreshID device that can accurately estimate the freshness of produce. The value of such a device, an imager or lens offreshness signatureslies in the ability to accurately gauge produce freshness, independent of its original harvest condition or its microbial contamination status. In addition, the FreshID toolbox including these indicators and molecular markers associated with freshness can be used in the process of breeding for delaying post-harvest degradation and extending the shelf life of vegetables and fruits.
Hyperspectral imaging of broccoli senescence to indicate the freshness of vegetablesThe value of hyperspectral imaging stems from its promise of being able to identify and distinguish between different materials (or chemical compositions) at a sub-pixel level. This is often accomplished through spectral unmixing analysis (that is, estimating the materials found in a scene and each pixel's percentage associated with each material). However, standard spectral unmixing analysis fails to use all available information about a scene and is generally limited to simple mixing models that do not take into account known physical realities associated with hyperspectral imagery (such as variability and locally constrained materials). Furthermore, unmixing of hyperspectral data is an ill-posed inverse problem and multiple solutions exist. Many methods have been developed to estimate solutions by constraining the solution space using sparse assumptions, geometrical constraints, and other approaches. These methods do not constrain the problem using scene-specific information to guide analysis to the right answer but, instead, constrain it based on broad assumptions.In this work, we have extensive knowledge about the objects being images and this knowledge can be leveraged to improve hyperspectral analysis and unmixing. Information such as time since harvest, nutritional content, maturity and firmness can be used to constrain the model used for unmixing and analysis. Specifically, we will develop a hierarchical Bayesian model and associated parameter estimation algorithms forSemi-Supervised Partial Membership Latent Dirichlet Allocation(sPM-LDA) for hyperspectral unmixing. The proposed sPM-LDA approach will leverage ancillary information to guide analysis, account for spectral variability, allow for localized materials/chemical compositions, and incorporate spatial information.RNA-Seq:Total RNA in samples ground in liquid N2will be extracted using TRIzol and RNeasy Plant Mini Kits (QIAGEN) and used for construction of a sub-library for multiplexing. The RNA-Seq for broccoli will be performed at the Interdisciplinary Center for Biotechnology Research (ICBR) at the University of Florida. A Biomek FXP Laboratory Automation Workstation with TRobot, which performs every aspect of liquid handling, will be used for library creation and sequencing. The platform is capable of walk-away automation of the preparation and sequencing of 200 RNA-Seq libraries a week using TruSeq RNA sample prep and NextSeq 500 system (Illumina). This equipment supports 2x 150-bp sequencing. One high output run result in ~400 million reads of sequencing.Data analysis of RNA-Seq:Sequence reads will be mapped to available genomes forBrassicaoleraceaandBrassica rapa(http://plants.ensembl.org/Brassica_oleracea&http://plants.ensembl.org/Brassica_rapa).For RNA-Seq data analysis, the Galaxy platform at UF Research Computing will be used. Workflow for read alignment will begin with fastq groomer, fastq QC, and fastq trimmer, then HISAT2 (daehwankimlab.github.io/hisat2/).Parameters will be adjusted for read lengths (typically about 150 bp for NexSeq500), but default settings will otherwise be used. Resulting bam files from HISAT2 will be analyzed for expression levels using Cufflinks, Cuffcompare, and Cuffdiff (http://bowtie-bio.sourceforge.net/index.shtml; Trapnell et al. 2012,http://cufflinks.cbcb.umd.edu/). Data will be extracted using the CummeRbund module (http://compbio.mit.edu/cummeRbund) in R (http://www.r-project.org)or blast2go (http://www.blast2go.com).Protein extraction and sample preparation:Proteinswill be extracted and purified from ground broccoli floret samples using a modified phenol extraction, followed by ammonium acetate-methanol precipitation. Eluted peptides will be ionized by a nano-electrospray source that integrates the LC-MS system. We will use the standard protocols provided by the UF-ICBR. Protein concentration in all extracts will be determined using the RC/DC™ protein assay kit and interfering compounds will be compensated for using the manufacturer's protocol. Bovine serum albumin will be employed as a standard. Proteins will be labeled using the stable dimethyl isotopes.Metabolomicsanalysis:We will prepare samples provided by UF-ICBR facilities to make compound volatiles. GC-MS-based metabolomics profiling and data analysis will be conducted by the ICBR. Integrated omics data enable the discovery of novel factors regulating the senescence and signaling pathways in broccoli.