Funded Seed Grants

This proposal aims to provide guidelines and computational tools to help plant and animal breeding programs decide when and how to invest in new phenomic and genomic technologies. To accomplish our goal, we will organize a webinar series to discuss and catalog costs and constraint parameters of modern breeding programs, optimize the UC Davis Strawberry Breeding Program's allocation of effort between direct phenotyping, High Throughput Phenotyping (HTP), and genomic evaluations, and build a general optimization strategy to help deploy new technologies.

We propose to develop a Plant Stress Ontology (PSO) to standardize a controlled structured vocabulary that is used to describe plant stress responses and adaptations. Plant stress refers to any environmental or biological factor that impairs the normal functioning and growth of a plant. These stressors can include abiotic factors such as extreme temperatures, drought, and soil salinity, as well as biotic factors such as pests, pathogens, mutualists and diseases. The typical hierarchical lists of plant diseases and stress, for example those provided by the plant disease bulletins, agriculture extension programs, and plant disease compendiums of the American Phytopathological Society (APS) are a rich source of information. However, they are difficult to cite, explore, and use for automation, exacerbating tracking of changes in the context of evolving information. Therefore a standardized PSO and its edited version tracking will provide the basic framework for developing a common language for describing plant stresses. The species agnostic reference PSO will provide a standardized way to describe and classify different types of plant stresses, their manifestation, measurements, observations, affected plant parts and growth stages, and curated images, phenotype data tables and known molecular interactions, as conceptualized by Walls et al. This will allow accurate description and comprehensive data collection and analysis. The PSO development and adoption will provide consistency in annotation and data collection in phenomics projects, enable open discussion on sharing information about plant stress responses observed/recorded by different research groups and improve interoperability among online databases. PSO knowledge graph (KG) will also help in unifying similar concepts and diverse vocabularies used in projects on agriculture extension [4], plant ecology, plant genetics, and plant breeding, as well as for training machine learning tools for stress detection. We will organize two week-long hands-on workshops by inviting plant pathologists, physiologists and ontology designers to help guide the discussion, build strategies and guidelines leading to development of first versions of the PSO ontology for biocuration of genes, QTLs and genome to phenome training data. Each workshop will fund participation of 10 experts through nomination and targeted crop representation. Additional open invitations will be available to experts.

Recent advances in Unoccupied Aerial Systems (UAS) and sensor technology are now making it possible to accurately assess overall crop health status with fine spatial and high temporal resolutions at a relatively low cost. When UAS is appropriately equipped with sensors, these platforms enable fast and accurate data collection throughout the growing season. For these reasons, the UAS-based High Throughput Phenotyping (HTP) system is becoming a standard tool in plant science research as it can provide more consistent phenotypic measures and seamless coverage of the whole experimental field. UAS offers an innovative opportunity to develop the HTP system for precision agriculture, including crop yield prediction (Ashapure et al., 2020), plant-level phenotyping (Oh et al., 2020), and crop precision management (Bhandari et al., 2021) for field-level production, to name a few. Accordingly, the number of research articles on the use of UAS for agricultural applications has grown exponentially, as shown in Figure 1. As plant scientists increasingly gain access to tools for collecting big UAS HTP data, there is a growing need to generate biologically-informative, quantitative phenotypic information from the collected geospatial data (Jung et al., 2021). The massive volume of geospatial data generated by the research scientists and lack of software packages customized for processing these data make it challenging to develop transdisciplinary research collaboration around these data. This project aims to address the disengagement between big data and agricultural research scientists by developing an open-source online platform for big UAS HTP data management. The online platform will serve as a virtual center for

1. Managing and visualizing massive volumes of UAS HTP data,
2. Promoting active discussion between scientists and engineers without geographical limitations, and
3. Communicating research findings to the public.

The proposed online platform is expected to serve as a one-stop shop for accessing and analyzing UAS HTP data. We will implement the FAIR (Findable, Accessible, Interoperable, and Reproducible) principle (Stall et al., 2019) as a fundamental architecture so that the online platform will cultivate a synergetic collaboration ecosystem between scientists and engineers by providing an open online platform for transparent communication. The open-source online platform will adopt a bottom-up approach to build a sustainable ecosystem in the long run. The initial phase of the project will work with crop research scientists of four major row crops: (1) corn (Dr. Tuinstra), (2) soybean (Dr. Wang), (3) wheat (Dr. Ibrahim), and (4) cotton (Dr. Bhandari). The smaller initial group will help us make progress quicker and demonstrate a successful project, which will provide a firm foundation to pursue a bigger grant from federal agencies such as NSF and USDA for a larger national center.

Technological developments across the field of genomics have resulted in what is commonly referred to as the democratization of genomics; that is, the combined decrease in sequencing costs and expansion in applications (transcriptomics, epigenetics, SNP data) have enabled scientists to apply genomics techniques to an increasingly broad range of species and situations. For example, successful genomics programs have produced high-quality genomes for a diverse range of agriculturally important species and associated resources, and utilizing these genomes is supported by USDA National Research Support Programs such as NRSP-8 Animal Genomes, NRSP-10 National Database Resources for Crop Genomics and the related AgBioData project. However, while we can rapidly produce more sequence-based data more cheaply than previously, the bottleneck in applying these techniques now lies in the application of bioinformatics to make sense of the data produced. For most researchers, it is now easier to generate genomic data than it is to manage and analyze the resulting data. Several initiatives have been used to develop accessible bioinformatics capacity (e.g., CyVerse, Galaxy, etc.) but these platforms still require more technical expertise than is commonly available to many scientists and tend to be software agnostic, making open-source software available without recommending or benchmarking specific workflows commonly used for agricultural genomics. Moreover, the agricultural genomics communities continue to amass multiple types of 'omics and genetics data at an increasing rate but there is limited re-use of this data. The agricultural communities' inability to repurpose existing data sets for gaining insight to novel questions represents a loss of value that is measured in the cost of data collection (and re-collection) and in the time and resources invested in collecting these data sets and to hinder the integration of genomics data sets for future genome to phenome analyses.

Unoccupied / unmanned aerial systems (UAS, also termed UAVs or drones) have unique advantages for advancing the phenome component of genome to phenome research (Shi et al. 2016, White et al. 2021, Jung et al., 2021; DeSalvio et al., 2022). UAS are becoming relatively inexpensive scientific instruments for leveraging the massive amounts of field-grown research plots (Bhandari et al. 2022, Adak et al. 2022; Rejeb et al., 2022; Jang et al., 2020; Haghighattalab et al., 2016) or animal and rangeland research (Abdulai et al. 2021, Gillian et al. 2021, Los et al. 2022). Unlike laboratory and genomics tools with expensive consumables, a field researchers' investigations can and have been leveraged into world class discovery with as little as a DJI Phantom 4 drone ($2500) and a computer ($5000); very accessible and scalable to those with limited resources. Based on successful knowledge and decision making (Guo et al. 2021), as well as anticipated future applications, UAS could become an important tool in every plant and animal field researcher's phenotyping tool-box within the next five years, but adoption will be stymied unless a community is built to share experiences and advance the technology. Many successful and potentially transformative case studies already exist. However, given accessible UAS technologies are recent, examples and analyses tools are new and largely exist in silos. For most researchers attempting to incorporate UAS into their research, education and extension, unknowns and perceived limitations exist as barriers. Other researchers use cases have overcome such challenges, but disparate publication, reporting and communities leave gaps in awareness. Even at the most basic level we experienced research published in the same species using the term 'UAV' was unaware of highly similar work published four years earlier because the term 'UAS' was used. In other cases, researchers in different regions, crops or species have developed a better or easier methods to use for the same task. Such levels of technical application and knowledge are considered scientifically unremarkable; thus details, tips and tricks that would allow a new investigator to get started are unlikely to be published in peer-reviewed publications. Other sources of this information have been developed and are available such as webinars (e.g. PhenomeForce, Plant Phenome Journal), manuals, tutorials, and software but remain incomplete, developed in silos and/or difficult to find and synthesize.

Phenotyping is a critical step for plant and animal research, both in practical applications such as the selection of superior individuals in breeding programs or for research applications like the precise dissection of the genetic architecture of important traits. One of the main challenges of phenotyping is the sheer amount of data that must be collected, analyzed, and stored. Traditional methods of phenotyping are time-consuming, labor-intensive, and subject to human error. Additionally, researchers are limited in the number of phenotypes they can collect in a growing season, often resulting in uncollected traits or reduced population sizes.

Imaging provides a solution to many of these issues, allowing researchers to quickly and easily capture large amounts of data with a standardized collection process to ensure accuracy and consistency. Modern image processing techniques can provide quantitative measures for complex traits and images can be asynchronously processed, allowing researchers to extract more detailed information from larger populations and reprocess data as new techniques are developed. However, many imaging approaches available to plant and animal scientists result in unlabeled data, complicating analysis or even resulting in unusable data.

Field Book is an open-source Android app for phenotyping used in the breeding community with more than 6,000 users worldwide. Multiple databases including BreedBase and the Breeding Information Management System have built in direct support for syncing with Field Book for data collection. The app allows users to collect trait data with an optimized user interface and features to ensure data integrity and continuity. While Field Book currently includes a feature that allows users to capture images linked to specific entries, this setup is currently limited in that: 1) it can only use the integrated device camera, restricting opportunities for fixed camera phenotyping that is common in plant and animal research and 2) it uses a nonoptimized file storage, slowing performance and restricting the ability to collect and utilize large amounts of data. The work outlined in this proposal aims to address both issues.

Applications of genomic prediction and genome-wide association (GWA) analyses in plant and animal agricultural species often face the problem of data sharing across multiple private and public institutions. This is particularly true for difficult to measure traits where several institutions are collecting phenotypic and genotypic data, but no single institution possess a dataset with enough individuals to obtain powerful GWA and accurate genomic predictions.

Thus, there is increased interest in methods that allow data integration and sharing while respecting privacy and intellectual property of each individual entity.

Several solutions have been used to circumvent the problem of data sharing in genetic studies. For instance, meta-analysis of GWA studies is commonly used by public-private consortia working on genetic epidemiology (Panagiotou et al., 2013); in this area our group has developed methods to perform GWA using results from multiple GBLUP genetic evaluations (Bernal Rubio et al., 2016). Likewise, meta-genomic-prediction has recently been proposed (Jighly et al., 2022). In meta-analysis each institution performs their own GWA and summary statistics from each of the studies are shared with a core group that performs the integration of results into a more powerful GWA or more accurate genomic prediction. Alternatively, monomorphic encryption (Blatt et al., 2020) has been used for genetic epidemiology to share data while protecting the privacy of each subject in the dataset, and maintaining marker-specific properties. This allows combining data and implementing tests of marker-phenotype association. Although these two approaches are promising and are already being used, there is still the need of methods that allow data integration without sharing data (either individual data or summary statistics) that may be sensitive.

Goats are the most adaptable and geographically widespread livestock species. Over one billion goats from more than 600 breeds are dispersed across the world, providing an important source of food for human from milk and meat. Despite the popularity and economic importance of goats, there are significant gaps in 1) the knowledge base of goat genome, pan-genome, and pan-epigenome, as well as their associations with goat phenome; 2) capacity such as expertise and resources dedicated to goat research; 3) adoption of advanced genomics technology by the goat industry. The overall goal of this project is to bring together experts from diverse disciplines and multiple research institutions to explore collaborative research opportunities in advancing goat genome to phenome research to improve goat sustainability and resilience to climate change.

The International Goat Research Center (IGRC) at Prairie View A&M University (PVAMU), an 1890 Land Grant Institution and prominent HBCU, currently houses more than 400 goats and has equipment and facilities to measure various economically important traits in goats. The PI and the Co-PIs at the IGRC have a breadth of expertise that is directly relevant to the goal of this proposal. Our collaborators at Agricultural Research Service (ARS) are the first in the world to develop a golden goat genome (ARS1). Our collaborator at the University of Idaho has recently received a grant from NIFA to develop the Ovine Pangenome. Our collaborator at Cornell University has led the African Goat Improvement Project to identify distinct genetic goat populations for future genetic improvement [6]. Our collaborator at the University of Pittsburgh has expertise in functional and computational genomics applied to immunology. The members of the working group have met on February 13, 2023, via a virtual meeting, and propose the following objectives/activities:

1. Create a network of researchers from diverse disciplines and multiple institutions to bridge the knowledge gaps in goat genome to phenome research. We will invite scientists to visit the IGRC, at PVAMU, to explore collaborative opportunities in goat genome to phenome research. After the on-site meeting, the group will host regular meetings to discuss current progress and future directions in goat (pan)genome, (pan)epigenome, and GoatGTEx research; explore multi-omics approaches to study economically important traits in goats; discuss potentials to study soil, plants, and animals in a systems approach; discuss key hurdles in implementing genomic selection in American dairy and meat goats to ensure food security in the US and the world. We will broaden our group to include experts from diverse disciplines, industry and goat producers, and will visit other's research centers to further explore collaborative research and develop joint grant proposals. We will write and publish a concept paper addressing the needs identified by the working group and give a presentation at an AG2PI workshop.
2. Build upon existing resources and identify emerging precision phenotyping technologies needed to improve goat sustainability and resilience. The team will exploit the rich phenotypic data accumulated at the IGRC and develop plans to incorporate genetic and genomic information to study economically important traits in goats. It will leverage existing resources, e.g., the GrowSafe (Vytelle) System, and identify emerging precision phenotyping technologies that can be used to improve goat sustainability and resilience. Our extension specialists will work with goat producers and consumers to better understand the needs and opportunities of this industry and translate our research outcome to benefit this underrepresented yet globally important species.

The 2023 NAPPN annual conference leverages the successes of the previous Phenome conference series (2016-2020) and the previous NAPPN annual conferences (2021-2022). The conference will assemble a range of sciences and technologies in engineering, agronomy, ecology, and plant systems biology from among academic, federal, and commercial entities to address the plant-environment interface. A key goal of the conference is to invite and support a new generation and a broad community of scientists from diverse scientific and cultural backgrounds. By participating in these types of exchanges attendees will accelerate the rate of advance in phenotyping by populating the field with innovative and neurodiverse thinkers. As part of our effort to diversify the plant phenomics research community, we request funding from AG2PI to support the participation of students, postdoctoral scholars and faculty members from minority serving institutions (MSIs) including Historically Black Colleges and Universities (HBCUs), Hispanic Serving Institutions (HSIs), and Primarily Undergraduate Institutions (PUIs). We have announced this year's conference (https://www.plantphenotyping.org/) and intent to provide support through emails and direct contact with faculty and administrators at MSIs.

The 30th Conference on Intelligent Systems for Molecular Biology (ISMB) is the premier international conference for computational biology and bioinformatics. Hosted by the International Society for Computational Biology (ISCB) this annual conference is expected to assemble over 2,000 participants from a range of biological and computational disciplines with interests in bioinformatics, computational biology, imaging, high throughput sequencing, AI/machine learning, and systems biology as applied to the life sciences. This year the conference will be a hybrid event with in-person activities held from July

Many breeders, researchers, and students lack the expertise necessary to deploy high-throughput phenotyping (HTP) to benefit cultivar development and research. To address this community-wide gap in capacity, we plan to host an in-person workshop that will engage attendees in hands-on, step-by-step procedures to deploy HTP and make meaningful use of HTP data. Training materials will be made publicly available, and attendees will gain skills to serve as an HTP resource at their home institutions.

The 2022 NAPPN annual conference leverages the successes of the previous Phenome conference series (2016-2020) and the virtual 2021 NAPPN annual conference. The conference will assemble a range of sciences and technologies in engineering, agronomy, ecology, and plant systems biology from among academic, federal and commercial entities to address the plant-environment interface. A key goal of the conference is to invite and support a new generation of scientists from diverse scientific and cultural backgrounds into this field in hopes that participating in these types of exchanges early in their careers will accelerate the rate of advance by populating the field with innovative thinkers. As part of our effort to diversify the phenomics research community, we request funding from AG2PI to cover the costs of participation for minority participants from minority serving institutions (MSIs) including Historically Black Colleges and Universities (HBCUs). This funding will cover registration and travel support.

The overall goal of this project is to enhance transdisciplinary learning with a major focus in digital agriculture. The specific objectives of this project include: Objective 1: Organize a summer training internship program for undergraduate students with background in animal science, crop science, and computer science: Undergraduate students from Departments of Animal Science, Plant Science, and Computer Science at Texas A&M University-college station, TAMU-Kingsville, and TAMU-Corpus Christi respectively will be enrolled as an intern during Summer 2022. Students will have opportunities to receive course credits for ANSC 494 internship course coordinated by Co-PI at TAMU. Similarly, student from TAMU-K will also receive the course credits for internship course.

Datasets obtained from precision dairy monitoring tools (pedometer) and Unmanned Aerial Systems (UAS) collected from cotton field will be used as two case studies. Students will learn about the data collection procedures in both systems, create a standardized data base, and utilize the database to train and validate machine learning models to predict disease events in dairy cows and yield prediction in cotton. The goal is to develop student competency of understanding data generating systems, the database, and utilize it to develop data-driven tools for precision phenotyping.

An inaugural cohort of three students will be selected for this 10-week program from each of the three representative campuses. Students will attend the weekly training sessions on best practices of handling, management and curation of big data obtained from the two systems. Students will be assigned tasks each week that they are required to complete individually. Each week will end with a debrief session where students reflect on the progress and plan to adjust the approach for the upcoming weeks. This presents the opportunity for shared learning as the cross disciplinary team learns from each other.

Objective 2: To develop a training manual on big data management in agriculture: A training manual on the best practices of database management utilizing the resources created during this internship will be developed and tested by the cohort. Based on the student experience and recommendation, final version of the manual will be published for use in future cohorts.

The overall goal of this proposal is to evaluate the ability of a recently proposed homomorphic data encryption method to address privacy or intellectual property issues that prevent data sharing and to enable adherence to and capitalizing on the benefits of the FAIR (Findable, Accessible, Interoperable and Reusable) principles for research data and industry data. A recent review of issues and methods related to safeguarding privacy of genomic data in human genetics is in Wang et al. (2022).

Advancement of precision land management technologies enables producers to manage the landscape with grazing animals to strategically improve ecosystem health and sustainability. Among the more novel of these technologies is virtual fencing (VF) - borders without physical barriers - to implement precision grazing management (Anderson 2007; Umstatter 2011). VF systems operate via GPS-enabled collars on each animal. There is a three-way interaction between the collars, a base station in the field, and a user interface (software) on a computer that allows users to 'draw' their pasture boundaries. These boundaries transmit to the base station (operated by cellular and solar), which 'pushes' the virtual fence instructions to the collars. Livestock are controlled within the virtual pasture with an auditory stimulus followed by an electrical pulse if the animal goes farther into the virtual boundary. The system is designed such that animals learn the association between the auditory cures and the electrical pulse and respond to the auditory cues alone.

Agriculture has the greatest footprint of any human activity, and much work has gone into improving its sustainability (Harwood, 2020). In modern conventional agriculture some hope to mitigate impacts/costs through optimization while in agroecology some hope to create holistic, resource-conserving methodologies for management. However, these two approaches to sustainable agriculture often come from different epistemological viewpoints; as a result, it is difficult both intellectually and practically to determine the best or even a good course of action in sustainable farming today (Jordan and Davis, 2015). While much work has gone into exploring complex cropping systems that provide more ecosystem services while producing the same amount of food, feed, fiber, and fuel as simpler systems (Tamburini et al. 2020), these systems are often idiotypic (Shaffer et al., 2000) and not transferable outside of the farms where they were trialed (Robertson et al., 2012). As computing has penetrated nearly all aspects of modern society (e.g., transportation, health and medicine, and human interaction), many have proposed to leverage computing to improve the sustainability and productivity of agriculture (Raghavan et al., 2016). We propose a way of merging individual farm-based solutions and accommodating different epistemological frameworks by borrowing tools from computer science---in particular, the notion of a state space (e.g., plant traits, cropping system) which can be explored by an artificial agent.

With the advancements of machine learning and artificial intelligence in digital agriculture, especially the precision agriculture phenotyping sensors and tools. There is a gap between HBCU (Historically Black Colleges and Universities) education, research, outreach, and the advances in precision agriculture phenotyping technologies. Florida Agricultural and Mechanical University (FAMU) is an 1890 land-grant institution (#1 Public HBCU by U.S. News & World Report) dedicated to the advancement of knowledge, resolution of complex issues and the empowerment of citizens and communities. As the land-grant arm of FAMU, the College of Agriculture and Food Sciences (CAFS), PI Chen's home college, plays a vital role in providing researched-based information and resources directly to Florida's farmers, individuals, producers, communities, and agri-businesses. FAMU CAFS Center for Viticulture and Small Fruit Research is recognized internationally for excellence in warm climate grape research and facilitator of outstanding academic programs for experiential learning and student training. Viticulture Center maintains the most extensive muscadine grape germplasm collection in the world and is serving as one of the five National Clean Plant Centers for Grapes. The Biological Systems Engineering (BSE), PI Chen's home program, is a branch of engineering which integrates agricultural, biological, chemical, and engineering sciences. The BSE program is one of the two ABET (Accreditation Board for Engineering and Technology) accredited BSE programs among the nineteen 1890 HBCUs in the U.S. Currently, there is a critical need for CAFS especially BSE program to develop education, research, and extension training on precision agriculture phenotyping tools.

Recessive lethal alleles exist benignly in breeding populations, until a sire and dam carrying them are mated. One quarter of the resulting pregnancies will be homozygous for the lethal allele and will result in an aborted pregnancy. Missed breeding opportunities are expensive. These recessive lethal alleles will increase in frequency within the population, distributed as heterozygotes, until ultimately manifesting themselves as lethal when two heterozygous carriers are mated. If it is possible to identify these lethal alleles, then farm managers can mitigate the problem by ensuring that two carriers are never mated to one another, thus boosting the likelihood of a successful pregnancy by 25% for any carrier.

This project seeks to develop an information hub for training/teaching agricultural researchers with an aim to facilitate application of single-cell functional genomics in quantitative trait loci (QTLs) mapping of agriculturally important traits. Integration of genetic variation data with cellular and molecular data has been used to map expression QTL (eQTL) or methylation QTL (mQTL) or chromatin accessible QTL (caQTL) linked to diverse phenotypes (Kumasaka et al., 2016; Volkov et al., 2016; Ciuculete et al., 2017; Benaglio et al., 2020a; Keele et al., 2020; Zhao et al., 2020). Such approaches have also been applied to map QTLs linked to traits of agricultural importance (Long et al., 2011; Liu et al., 2020; Kushanov et al., 2021; Yuan et al., 2021). However, these studies have been performed with functional genomics data derived from bulk tissues that cannot determine if the phenotype is influenced by specific cell types in the tissue. Recent spur in single-cell RNA sequencing (scRNA-seq) and single-cell Assay for Transposase-Accessible Chromatin sequencing (scATAC-seq) have generated new opportunities to integrate genetic variation with changes in gene expression and open chromatin profiles to identify single-cell eQTLs and caQTLs (Benaglio et al., 2020b; van der Wijst et al., 2020; Neavin et al., 2021). These methods have great capabilities to untangle cellular and molecular links to major phenotypic traits such as crop yield, animal production and plant resistance to insect pests to sustain agricultural productivity (Cole et al., 2021; Tripathi and Wilkins, 2021; Zhang et al., 2021; Zhu et al., 2021; Nyyssölä et al., 2022).

The resurgence in sustainable farming practices in recent years is driven mostly by interests in improving soil health, nutrient cycling, and carbon sequestration. However, most of the research has focused on utilizing annual cover crops, which are often terminated at the end of the season, and the benefits of alfalfa (Medicago sativa L.) in cropping systems have been largely overlooked. Due to its perennial nature, alfalfa can improve soil structure, decrease erosion, and increase carbon sequestration in soil. Increased utilization of alfalfa will not only help to reach ecological goals, but it will also help in improving wildlife habitat and biodiversity, while providing a highly nutritious feedstuff for livestock.

Lameness presents a major animal welfare concern and is a significant economic burden for the livestock industry. For example, lameness costs the dairy industry alone around $52 million a year. Current methods of assessing lameness, conformation and locomotion phenotypes are often plagued by a lack of repeatability and accuracy, yielding heritability values for lameness of just 0.01 and 0.22; indicating need for a more accurate phenotyping approach for locomotor traits. Visual assessment, the most common approach, lacks repeatability [6], accelerometer and gyroscope methods alter natural gait patterns, and reflective 3D markers are not feasible in less tractable livestock systems where application of reflectors and utilization of multiple-camera detector arrays is impractical and costly. StepMetrix technology in cattle has documented stance time and ground reaction force but has yet to consider diverse locomotor phenotypes indicative of lameness like back posture. This project will utilize a published machine learning package (DeepLabCut, DLC version 2.2b7) in combination with a custom gait analysis pipeline to produce quantitative locomotor phenotyping protocols specifically for livestock. Previous work demonstrated the deep neural network (DNN) approach employed by DLC can label landmarks on an animal with the same accuracy as the human eye but in far less time. For example, the pilot project described below would take one full-time operator about four months of continuous work to label the 77,000 frames of data, but only a few hours by applying the DLC machine learning approach.

The analysis of how genome information creates phenotypes at the single cell level, the fundamental unit of biology, is a powerful approach for understanding genome function, and is rapidly becoming the gold standard for human genetics research predicting phenotype from genotype. The multicellular complexity of plant and animal agricultural species limits our understanding of the regulation and organization of their genome, and the expression patterns of their genes in each cell composing these species. To make the enormous promise of single-cell (SC) genomics a reality for the agricultural genome to phenome community, we need to develop Findable, Accessible, Interoperable, and Reuseable (FAIR) SC data resources and informatic tools for storing, sharing, and analyzing such data that is currently accumulating in crop and livestock research groups. We believe this Enabling seed grant proposal addresses topic areas #1 and 2 in the AG2PI RFP. The lack of FAIR SC data and the computational skills required for researchers to use such data currently prevents the adoption of this powerful method within the AG2PI community.

There is a need to characterize the variability of the dairy cow heat stress response and to identify non-invasive phenotypic indicators of heat stress that can be automatically detected using existing technologies. We aim to

1. Characterize regional variability in productivity responses to heat stress
2. Identify and collate the needed phenotyping data to characterize the heat stress response
3. Develop strategies for integrating disparate data types from animal monitoring systems to create a relational database of non-invasive phenotypic indicators of heat stress

Animals in extensive production systems are faced with many environmental challenges which may impact their ability to perform. We will use GPS units as a precision agriculture management tool to make land-use management decisions to maintain a healthy ecosystem, track animal behavior, and develop novel trait definitions for individual performance in a rangeland landscape. We expect GPS collars will provide a means to address the economic and environmental costs of an extensive sheep operation by providing quantitative measures of animal behavior in a rangeland environment.

Plant and animal communities are accelerating the creation of functional genomics data. To capitalize on these investments, better methods and data standards are needed to integrate disparate datatypes, predict regulatory elements and glean new insights. We propose to bring together experts in plant and animal functional genomic data reuse and sustainability for a workshop and on-going discussion groups to identify and prioritize shared needs for data re-use tools to link genotype to phenotype.

While the technical aspects of data integration are achievable, the metadata collection required for robust meta-analysis of G2P and G2E studies, especially for non-model plant systems, remains a hurdle. We will develop a fully FAIR data submission module (TPPS) that can be implemented by all Tripal plant databases, integrate the WildType mobile application to collect trait data for landscape-based studies, and train the scientific community from biocurators to Tripal database administrators.

Currently, the biggest challenge for underrepresented and non-expert users is to have access to AI techniques through a more user-friendly interface to perform basic AI tasks without requiring extensive expertise in the corresponding areas. In this project, we will develop an open-source software to democratize the access to AI techniques. Such software will be used to perform image classification by training new and customized deep learning algorithms on datasets provided by the user. Our goal is to create more accessible user interfaces so that coding ability is not a barrier.

Event-based phenotyping analysis refers to the timing detection of the important events in a plant's life. This research will develop deep learning-based algorithms for emergence timing detection and growth tracking of seedlings using time-lapse image sequences, and detecting flowers and fruits from multi-view images to compute reproductive stage phenotypes. We will publicly release a benchmark dataset to develop and evaluate the algorithms. A software tool called iPlantSeg+ will be released to allow non-experts to perform segmentation and compute common phenotypes.

We propose using a sub-$5 targeted gene expression approach as a molecular phenotype in beef cattle. This proposal aims to computationally identify 500 high-information genes that will be assayed in ~1,500 beef stocker calves. We will explore the utility of using these expression counts to predict future calf performance & health outcomes and as a latent phenotype. While we focus on beef cattle, we expect that this technology could be applied across species and genome to phenome applications.

Identifying genetic adaptation in non-model systems (NMS) often cannot be sufficiently addressed using standard marker assisted selection (MAS) methodologies. While MAS is indispensable for the selective breeding of traits, NMS have not been able to take advantage of advancements in MAS due to the high overhead and imperfect data they often face. NMS contend with imperfect pedigrees, smaller populations, missing phenotypic/genotypic information, and complex interacting genetic components. Machine learning (ML) methods have shown to be tolerant of these biases and offer an alternative means of providing markers for genome selection. In spite of the potential for ML to vastly improve MAS in NMS, little is available by way of tools for researchers to utilize for breeding programs. We plan to address these problems through

1. Development of an effective ML-based algorithm tailored for genome selection
2. A simple to implement tool for use by those that are familiar with breeding/MAS

This tool will identify variant locations that are contributing to phenotypic variation of a dataset without adding to user workload by utilizing high throughput genotypic and phenotypic information that is common to a MAS breeding program. The selected variant sites will be utilized as the basis of genetic markers for population screening and selection towards the improvement of germplasm. This methodology would reduce the resources needed for MAS in non-model crop species or those with complex phenotypic targets.

The main goal of this project is to develop online educational material for managing and sharing UAS based HTP data using public cloud services. Although the importance of the FAIR principles in data-intensive science has been addressed multiple times, little attention has been paid to management and sharing of the big geospatial data generated from the UAS imageries yet. We propose to develop online educational material to provide tutorials on how to share the geospatial data products generated from the UAS data with the general public as web services.

The use of scale-invariant properties can improve our understanding of genome to phenome associations. This project uses the first principles of biophysics to develop cross-scale correlations between photosystem II related genes and drought phenotypes for agricultural species. We will perform a comparative genomic study searching for changes in the sequences to be imported into protein crystal structures for molecular modeling. Modeled water affinity across species will be correlated to existing phenotypic information to build associations for the ability of plants to grow and strive under water limitations.

Despite years of research, there is not a clear understanding of how genotype contributes to phenotype of beef cattle. Our long term goal is to determine how underlying genetic differences present between cattle of different breed types translate to differences in animal performance, carcass quality, environmental impact and therefore economic viability. The overall objective of this proposal is to gather preliminary data to better understand how genetic differences relate to economically important traits and establish a resource for future genome and phenome comparison and manipulation.

The AG2P Initiative strives to connect genotype to phenotype in many environments. A growing source of phenome data in agricultural research across animals and plants are Unoccupied Aerial Vehicles (UAV). In this proposal we build a foundation to empower more researchers to use UAS. In the objectives, we will complete an international survey of agricultural animal science and plant science researchers of UAV use and build a community-informed web resource to provide instructional information and benchmarking tools to support the growing number of UAV users.

Expected Outcomes and Deliverables: In summary, this project will provide the foundation to accelerate high-throughput phenotyping in agriculture to support the mission of AG2PI. A survey will assess how UAV imagery has been applied to plant and animal sciences in agriculture and identify obstacles common among research groups that will pinpoint potential solutions. Simultaneously, developing standardized best practices will support UAV adoption and reduce barriers to entry. These efforts will be shared through workshops, videos, conferences, and manuscripts.

The AG2P Initiative strives to connect genotype to phenotype in many environments. A growing source of phenome data in agricultural research across animals and plants are Unoccupied Aerial Vehicles (UAV). In this proposal we build a foundation to empower more researchers to use UAS. In the objectives, we will complete an international survey of agricultural animal science and plant science researchers of UAV use and build a community-informed web resource to provide instructional information and benchmarking tools to support the growing number of UAV users.

Sustainably supporting worldwide food production is a wicked problem of immense scale and complexity. As such, there is an urgent need for novel ideas and technological innovations in the agricultural research. One strategy for rapidly infusing new ideas into existing knowledge networks is by making these person-centric networks more diverse, data more accessible to relevant stakeholders, and by infusing current practices with the kinds of Ethical, Legal, Social, Ecological, and Economic (ELSEE) considerations that will transform agricultural genome and phenome research practices. Bringing underrepresented groups to the table is one way to infuse AG2PI with new ideas. Improving and expediting knowledge transfer (e.g. data sharing), more effectively communicating research findings to the general public, policy makers, and funding agencies, and developing new science practices, can also help AG2PI to achieve sustainable genetic improvements. We propose to advance the aims of the AG2PI by conducting social science research to encourage cross-fertilization of AG2P data and ideas and motivate agriculture focused analysis from an ELSEE perspective. We will also create human-centered

The main goal of this project is to form public-private partnerships to expand meaningful interactions between industry and the public sector. Outcomes will include:

1. Enhancing graduate student training through the use of current real-life AG2P project scenarios and datasets;
2. Generating public educational resources including datasets and code for use by other scientific communities for AG2P training;
3. Serving as a model and test bed for PPP engagement and seeding meaningful ongoing collaborations in interdisciplinary groups.

Implementing machine learning approaches to connect genotype-phenotype has been hindered by the lack of available, labeled training datasets. To overcome this limitation, we will use geometric modeling and object reconstruction, using point cloud data, to develop simulated datasets of organismal development. These simulated growth models will generate labeled datasets which can be used by machine learning algorithms as training data to study the complexities of phenotypic diversity. Moreover, we will develop and test a system for illumination estimation of the virtual crops.

Aim 1: We will provide a large, labeled dataset of point cloud data for sorghum plants with varying precision for ML and evaluation by the plant science community. We will develop the generative procedural model of sorghum that will be parameterized by branching angles, plant age, etc.

Aim 2: We will provide a highly parallel path tracer simulating sorghum illumination at a photonic level at varying wavelengths. This simulator will be expandable for further BRDF and BTDF values and will be parameterizable for varying longitude and latitude.

1. Novel 3D simulations of sorghum grown at high-density with plant parts segmented & labeled,
2. Plant and organ labeled simulated point cloud data for sorghum at high planting densities,
3. ML models using simulated segmented plant data tested on greenhouse and field LIDAR data,
4. Improved understanding of light interception and interaction dynamics with plant canopy,
5. Engagement with animal researchers to assess how to extend approaches to animal science,
6. Hosted AG2PI workshop to engage with plant and animal scientists and expand community.

In this project, we will organize competitions to engage with the broader machine learning community to produce models that can answer phenomic questions, using curated datasets from the Department of Energy ARPA-E Transportation Energy Resources from Renewable Agriculture Phenotyping Reference Platform (TERRA-REF) program [Burnette et al., 2018]. The TERRA-REF program aimed to transform plant breeding by using remote sensing from a state of the art field scanner gantry system, seen in Figure 1 (top) to increase the speed at which plant traits can be measured. The field scanner includes a number of sensors, including a millimeter resolution laser 3D scanner, high resolution stereo-RGB cameras, multiple hyperspectral sensors, and a thermal camera, among others (example data products are shown at the bottom of Figure 1). Over the course of several seasons, this sensor collected over a petabyte of sensor data for bioenergy sorghum lines, their corresponding genetic data, a large volume of ground truth measurements of plant phenotypes and growing conditions, and a baseline set of algorithmic approaches for extracting phenotypic data.

We propose that structured machine learning contests are an excellent way to share data and ensure that the desired scientific questions are actually the ones that are answered. Contests, where a specific problem is shared, training and testing data are provided, and a specific evaluation protocol is defined, are a frequent and popular means of advancing results to specific questions within the machine learning community. For example, the Fine-Grained Visual Classification community hosts annual contests on difficult visual recognition problems, such as the iNaturalist contest to recognize different species in image data, which had 1,477 submissions last year from 249 different competitors [Van Horn et al., 2018]. These communities are hungry for well organized datasets with specific scientific questions.

Recent developments in genomics and the implementation of new technologies in agriculture have enabled a new research horizon in the field (Harper et al., 2018, Morota et al., 2018). The quality and quantity of genomic data, including microbiome, gene expression, high density SNP markers, sequence data, among others, have great potential to improve both plant and animal science enterprises. However, with these new developments a new challenge has arisen: practitioners must be able to manage the large data sets that result from these technologies – a skill that has traditionally is not been taught during the training of agricultural scientists (Eisen, 2008).

The goal of the Seed Grant is to catalog the available resources and resource gaps in data science to support the Agricultural Genome to Phenome (AG2P) initiative and to outline solutions to fill the gaps. We will develop surveys to identify how aware students and researchers are of the available resources. Additionally, we will create an online repository linking to available training materials in both plant and animal agricultural data science. In this repository we aim to provide the community a unified access point to information about workshops, seminars, online and in-person classes, and course curricula for different career stages. We will prepare a white paper describing our findings based on the survey and the catalog, focusing on how to advance data science education in AG2P. We will fund a graduate student to carry out this work. This student will work in the laboratory of PI Fragomeni with additional support from our team of investigators. We will use the results of the seed grant as preliminary data to apply for a large educational project to develop solutions to the needs identified in this project.

The overarching goal of the Cattle Genome to Herd Phenotyping for Precision Ag initiative is to exploit new phenotyping technologies and high throughput genomics to improve cattle productivity and profitability. To accomplish this, we will establish a network of researchers from a variety of disciplines and agencies to facilitate generation and implementation of next generation phenotyping technologies in cattle and ascertain existing resources.

Expected Outcomes and Deliverables: PIs McKay, Hagen, Schnabel and Murdoch expect to form a CG2HP working group consisting of scientists from a variety of disciplines (i.e. nutrition, physiology, engineering and economics). Scientists will represent leaders from universities, government, industry and breed associations. This working group will be expected to attend an initial virtual meeting in September and meet in person in San Diego in January 2022. A CG2HP concept presentation will be given at PAG and a concept paper addressing the needed phenotyping data and technologies necessary to implement CG2HP in cattle will be generated and published. Further directions and future action items will be discussed, including future funding opportunities. Additionally, a presentation will be given at an AG2PI conference or workshop.