Source: Central State University submitted to NRP
GENETICS AND BREEDING OF MITE BITING BEES FOR RESILIENCE TO VARROA MITES
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
COMPLETE
Funding Source
AFRI COMPETITIVE GRANT
Reporting Frequency
Annual
Accession No.
1022433
Grant No.
2020-67014-31557
Cumulative Award Amt.
$180,000.00
Proposal No.
2019-06121
Multistate No.
(N/A)
Project Start Date
Jun 1, 2020
Project End Date
May 31, 2024
Grant Year
2020
Program Code
[A1113]- Pollinator Health: Research and Application
Recipient Organization
Central State University
1400 Brush Row Rd.
Wilberforce,OH 45384
Performing Department
1890 Land Grand Research
Non Technical Summary
Honeybees are the primary source for pollination service to produce fruits, vegetables, and crops in the U.S. However, they face severe colony losses since 2007. Improving honey bee resistance to Varroa mites is critical to maintaining sustainable apiculture. However, there are few research efforts to investigate the resilience of behavior against Varroa mites. Our preliminary research has shown that the mite biting behavior in Ohio feral bees and Purdue University breeding stocks are higher than commercial stocks. We have identified a group of candidate colonies for breeding in Ohio. Our project aims to 1) Investigate the difference of evolutionary morphology and mechanism of mandibles in Midwest mite biting bee populations (Apis mellifera) and Eastern Honeybees (Apis cerana); 2) identify transcriptomic profiles and genetic markers for mite biting behavior, and 3) apply knowledge of in-situ techniques and a phone application to increase the accuracy of screening MBB bees, and improve honeybee health by genetics and breeding mite-resistant bee populations. This study will be the first innovative attempt to improve the resilience trait of European Honey Bees (Apis mellifera) against Varroa mites in Ohio and second in the Midwest. We expect our new bee program to provide a long-term solution to mitigate the 60% honeybee losses in Ohio and other Midwest regions. Central State University (CSU), Ohio's only public HBCU and the newest 1890 land-grant institution in collaborating with Purdue University submits this proposal which will enhance CSU's research capacity in pollinator health and sustainable apiculture, and also train minority students to join the future USA's sustainable agriculture workforce.
Animal Health Component
50%
Research Effort Categories
Basic
50%
Applied
50%
Developmental
0%
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
30330101080100%
Knowledge Area
303 - Genetic Improvement of Animals;

Subject Of Investigation
3010 - Honey bees;

Field Of Science
1080 - Genetics;
Goals / Objectives
Our central hypothesis is that worker bees in more resistant colonies (Oh feral colonies and Purdue MBB) evolve mandibles that are better at biting. We aim to investigate the evolution underlying the morphology and genomics of mandibles between mite susceptible (commercial bees) and mite resistant colonies (feral, Purdue, and A. cerana).Our Objectives are: 1) Compare the evolutionary morphology of mandibles among mite resistant populations of Apis mellifera and Apis cerana to elucidate the potential behavioral microevolution of mite biting behavior; 2) Investigate the genetic basis of mite biting behavior via transcriptomic profiles of bee mandibles, and 3) Apply knowledge of in-situ techniques and a phone application to increase the accuracy of screening MBB bees, and improve honeybee health by genetics and breeding mite-resistant bee populations.
Project Methods
Materials and Methods?To compare the morphology of mandibles, we will collect 30-50 pollen forager bees from 4 different populations, 1) commercial package bees (A. mellifera), 2) Purdue MBB (A. mellifera), 3) Ohio feral colonies survived/established for at least 3 years over winter (A. mellifera), and 4) A. cerana in Beijing, China. Mandibles will be compared using the microCT scanning at Ohio State University. For statistics, please see the Data Analyses section.For the genetic basis of mite biting behavior, Illumina RNA-seq analysis between Purdue MBB, commercial/package bees, Ohio feral bees, and Apis cerana population. We will collect the 8-day-old marked bees to make sure they are all the same age. Total RNA will be extracted from both dissected tissues of the 3 sampled individuals per treatment group and convert it into a barcoded Illumina™ sequencing library for multiplexed high-throughput sequencing on an Illumina MiniSeq. We will obtain 12 samples (4 populations x 3 individuals per population) for transcriptomic analysis. Reads will be quality-filtered and aligned to the honey bee reference genome (Apis mellifera DH4, NCBI - ID #: 9555). Whole-genome sequencing will be carried out using whole-genome DNA extraction. Libraries from 12 samples (4 populations x 3 individuals per population) will be prepared for Illumina MiniSeq sequencing. We will identify variations of the DNA sequences in the form of single-nucleotide polymorphisms (SNPs) for selection and breeding.To identify more MBB colonies in Western and Central Ohio for the future, we will collect mite data in colonies within 48hrs window (in the summer and fall) with beekeeper associations in nearby counties (Greene, Clark, Brown, Montgomery, Miami, and Shelby) of Central State University. The data of mite biting behavior and mite numbers will inform beekeepers about the science behind the mite resistance behavior. For more statistics, please see the Data Analyses section.We propose to develop a smartphone application (both in iOS and Android systems) to classify mutilated versus intact mites in the apiaries. Our innovation will significantly improve the efficiency of mite counts, the accuracy of damaged mite counts, and reliable data recording using a phone device that farmers have easy access to. In addition, we plan to promote this application to other beekeepers nationally if they are interested in measuring the mite biting behavior of their own colonies for breeding efforts.We will screen possible 30 established feral colonies from swarm traps and 30 commercial/package colonies for mite biting behavior. Our team will use samples to test the accuracy of the classifier. At the same time, we will use more field data to optimize the model. We will develop a customized neural network to identify damaged mites using transfer learning. Transfer learning is commonly used in deep learning applications. Fine-tuning a network with transfer learning is usually much faster than training a network with randomly initialized weights from scratch. The pre-trained features can be quickly transferred to a new task using a much smaller number of training images, while the robustness of the pre-trained network is maintained. State-of-the-art pre-trained networks such as MobileNet and Inception v3 will be fine-tuned in this project. Once the customized neural network has been created, an easy-to-use android app will then be developed for the on-site classification tasks. When the app is ready, researchers will create a registry and an open database online, which all beekeepers will have access to. We will submit our app to the Apple iTunes store and Android app store for approval. Results will be presented at the national/regional/local Bee Research Conferences and all of our extension activities with exposure to beekeepers.Data Analyses. For Obj. 1: The volumetric datasets generated by the microCT scanner HeliScan will be analyzed by a specialized open-source software called Tomviz. High-quality and interpretable visualization will be extracted from scanned 3D structures (Levin et al., 2018). Figures will be produced for scientific publication. Test for the difference of morphology will be carried out by analysis of variance among three groups (Apis cerana, MBB, and package bees) with at least triplicates. For Obj. 2: RNA-seq analysis will be carried out in Gene-expression data will be computed. Differentially expressed genes will be identified with the DESeq2 package in "R". For Obj. 3: We will initiate the project by developing the algorithm for the smartphone application in year 1. One undergraduate student will assistant in this project. We will collect mite images, including uninjured and injured mites as a database. We customize several networks using well-known pre-trained networks using transfer learning. We will test a generalized phone lens for both Apple and Android phones for more mite pictures.

Progress 06/01/20 to 05/31/24

Outputs
Target Audience:Target audiences are undergraduate students at Central State University, individuals, and groups of beekeepers and communities in Ohio. Students at Central State University are racial and ethnic minorities who are African-Americans, most of which are socially, economically, or educationally disadvantaged. We have delivered science-based knowledge of honey bee grooming behavior to students, beekeepers, and stakeholders through formal or informal educational programs. We created a new class as formal classroom instruction at Central State University, created new laboratory instructions on the grooming assay; developed curriculum and innovative teaching methodologies for bee biology and beekeeping; created summer research internships for CSU students. We also held online webinars and workshops to talk about honey bee queen biology with CSU extension team. Changes/Problems:It took so long to hire a postdoctoral researcher to continue the project, and we eventually ran out of time for this project. What opportunities for training and professional development has the project provided?We have provided many research training opportunities for CSU students, who are African American students as underrepresented population. They received training in scientific research, leadership training, technical knowledge and skills in beekeeping, molecular biology, apiculture, ecology, and evolution. How have the results been disseminated to communities of interest?We have attended the American Bee Research Conference 2022and 2024, which were also joint meetings with the American Beekeeping Federation, and let American beekeepers know our research results. We also presented our research at Ohio State Beekeeping Association conference, in Kentucky queen bee breeders association, Warren county beekeepers association, brown country beekeeper association, greene county beekeeper association in the state of Ohio. What do you plan to do during the next reporting period to accomplish the goals? Nothing Reported

Impacts
What was accomplished under these goals? Our central hypothesisis that worker bees in more resistant colonies (Oh feral colonies and Purdue MBB) evolve mandibles that are better at biting. We aim to investigate the evolution underlying the morphology and genomics of mandibles between mite-susceptible (commercial bees) and mite-resistant colonies (feral, Purdue, andA. cerana).We have compared the morphology and shapes of mandibles among mite-resistant populations ofApis melliferaandApisceranato elucidate the potential behavioral microevolution of mite biting behavior. When the total number of mites from the three populations of EHBs (commercial colonies, OH feral colonies, and KY feral colonies) were compared, no significant difference was detected from the regression analysis, which indicated a similar mite population distributed among all the bee colonies (Fig. 2). To compare the mite-biting behavior of the three populations (commercial colonies, OH feral colonies, and KY feral colonies), we used the ratio of damaged mites to total mites in each colony. In total, there were of 2518 mites from 58 commercial colonies and 4981 mites from 97 feral colonies. The regression analysis showed that the KY feral bees A. m. mellifera (ratio mean ± SD = 33.61% ± 0.13%, transformed ratio mean ± SE = 0.57 ± 0.026) displayed 34% higher mite-biting behavior than did the commercial bees (mean ± SD = 25% ± 0.11%, transformed ratio mean ± SE = 0.47 ± 0.015). The OH feral bees A. m. ligustica (mean ± SD = 37.26% ± 0.11%, transformed ratio mean ± SE = 0.60 ± 0.012) displayed a 49% higher rate of damaged mites than did the commercial colonies (mean ± SD = 25% ± 0.11%, transformed ratio mean ± SE = 0.47 ± 0.015; see Fig. 3 and Supplemental Material). The OH feral and KY feral bees were found to have similar mite-biting behavior (Q = 2.37). The different bee populations appeared to be a significant predictor of mite-biting behavior. To explore the effects of the year, we performed a regression analysis from fit models by using the standard least squares personality. The construct model effect was the year the samples were collected. Bee colony numbers were N = 20, 28, 57, and 50 for the years 2020, 2021, 2022, and 2023. The year appeared to be a nonsignificant predictor of mite-biting behavior (p = .91; Fig. 4). For the mandible observations, the locations of the colonies were inAmerica and Asia. We compared four parameters (long edge, short edge, height, and the span of the spine area)in the morphology of worker bee mandibles in four different bee populations (A. cerana, gAHB A. m. scutellata-hybrid, OH feral A. m. ligustica, and package commercial colonies A. m. ligustica). To provide details of the mandible structure, in the Supplemental Material we have included a 3-D movie showing the structure. For the long and short edges, commercial bees showed a significantly greater measurement than did the other three groups. For the comparison of heights, A. cerana showed the greatest height, and the OH feral A. m. ligustica showed the least. We found no difference in the span of the spine area among all the groups. To create a mobile application, a number of mobile-sized CNNs were considered in this study, which included ResNet50 (He et al., 2016), MobileNet v2 (Sandler et al., 2018), and EfficientNet B0-B4 (Tan and Le, 2019). Each of these networks also had a TensorFlow Lite version that could further increase the inferencing speed for phones, tablets, and embedded devices with limited computational power. A default parameter setup in Keras was used to retrain each neural network. We used 0.01 for the learning rate and 32 for the batch size, and we ran 10 epochs with 16 iterations in each epoch. When these parameters were changed, we observed no significant improvement in this experiment. When the same programs were run multiple times, a 2%-7% accuracy difference could occur because of random split of the images. A performance comparison is shown in Supplemental Fig. 2. The experimental results suggest that these pretrained networks were transferable to intact or damaged mite classification with desirable performance, considering the fact that a clear distinction between an intact or healthy mite and a damaged mite is sometimes difficult to tell even by the human eye. The best accuracy was achieved at 93.9% when using EfficientNet lite version B3 and B4, which is one of most accurate and efficient neural networks developed recently. ResNet50 achieved 80.3% accuracy, and MobileNet v2 achieved 87.9% accuracy, whereas the accuracy among EfficientNet B0-B4 varied from 83.3% to 93.9%. Greater accuracy might be achieved via a more complex neural network with the sacrifice of speed. It is safe to say that the experimental results proved the concept that intact mites and damaged mites could be distinguished by using artificial intelligence and that this technology could be used to assist bee farmers in counting damaged mites from each colony by using an average smartphone. When a precise count is not necessary to compare different colonies, a classification accuracy of approximately 90% is realistically acceptable. The application is available for download and to test in our cloud using Android devices with developer mode. This study showed that the mite-biting behavior in the OH feral colonies (A. m. ligustica) and the KY feral colonies (A. m. mellifera) was higher than in the commercial colonies. In addition, we provided evidence of the possibility of morphometric identification of differences in bee mandibles across different populations of Asian Honey Bees A. cerana, OH feral bees A. m. ligustica, the gAHB A. m. scutellata-hybrid from Puerto Rico, and commercial colonies A. m. ligustica. In addition, we developed a new mobile application utilizing machine learning and artificial intelligence to identify damaged mites via image classification. Mandibles serve various essential functions among worker bees. They are utilized in defensive behavior against parasites (Papachristoforou et al., 2012; Ruttner and Hanel, 1992). Through morphometric identification, we discovered morphological differences in three major parameters of bee mandibles (long edge, short edge, and height) among A. cerana and A. mellifera, including in two subspecies (A. m. ligustica and A. m. scutellata). On the basis of data on the long and short edges of mandibles, feral bee and gAHB bee mandibles tended to be similar to those of A. cerana and significantly smaller than those of the commercial bees, which indicates a possible macroevolutionary change in A. mellifera when they develop mite resistance. Since 1987, feral or wild colonies have possibly developed more mite resistance by increasing their efficiency of mite removal through natural selection (Hunt et al., 2016; Smith et al., 2021). We carried out transcriptomic and isoform analyses to investigate the genetic basis of mite-biting behavior by sequencing 15 pairs of workers'mandibles from either commercial bees (low biting), and breeding stocks (high biting). The postdoctoral researcher did not finish the data analysis and left the position. Currently PI is working on the data analysis for a publication. We appliedknowledge of machine learning in image classifications with co-PI in computer science and students from computer sciand a phone application to increase the accuracy of screening MBB bees, and improve honeybee health by genetics and breeding mite-resistant bee populations.

Publications

  • Type: Journal Articles Status: Published Year Published: 2021 Citation: Morphological changes in the mandibles accompany the defensive behavior of Indiana mite biting honey bees against Varroa destructor J Smith, XL Cleare, K Given, H Li-Byarlay, 2021, Frontiers in Ecology and Evolution, 243
  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2022 Citation: Mandibles from mite-biting bees: a multifaceted approach. M. C. Farrell, H Li-Byarlay, 2022 International Branch Virtual Symposium, Entomological Society of America.
  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2021 Citation: Potential mandible differences in the transcriptome among different bee stocks: A multifaceted approach. M. C. Farrell, H Li-Byarlay, 2021 Entomological Society of America Annual Conference (virtual)
  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2021 Citation: Comparison of grooming behavior between two honey bee species: Apis mellilfera and Apis cerana. H. Belton, H. LiByarlay, 2021 Entomological Society of America Annual Conference (virtual)
  • Type: Other Status: Published Year Published: 2021 Citation: Integrative Research to Study Social Honeybees: Genes, Behavior, Brain, and Health. H. Li-Byarlay, 2021. Entomology Seminar, Michigan State University
  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2023 Citation: Gene Expression Analysis and the Grooming Behavior in Different Ages of Worker Bees. H. Li-Byarlay 2023. Kentucky Queen Breeding Association Workshop
  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2023 Citation: Mandible changes and grooming behavior between two honey bee species. H Li-Byarlay, 2023 International Branch Virtual Symposium, Entomological Society of America.
  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2023 Citation: Grooming behavior in two honey bee species. H Li-Byarlay, 2023 International Conference on Pollinator Biology, Health, Policy. The Penn Stater Conference Center, June 3 - 6, 2023, University Park, PA 16802.
  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2023 Citation: Genetics and breeding of mite biting bees for resilience to Varroa mites. H Li-Byarlay, 2023 USDA Pollinator Health PD Meeting (virtual). 9/28/2023
  • Type: Other Status: Published Year Published: 2024 Citation: Genes and Genomics underlying the Social Behavior of Honey Bees. H Li-Byarlay, Biology Seminar, University of Florida, Gainsville, FL. 1/9/2024
  • Type: Journal Articles Status: Under Review Year Published: 2024 Citation: Biting behavior against Varroa mites in honey bees is associated with changes in mandibles, with tracking by a new mobile application for mite damage identification


Progress 06/01/23 to 05/31/24

Outputs
Target Audience:Target audiences are undergraduate students at Central State University, individuals, and groups of beekeepers and communities in Ohio. Students at Central State University are racial and ethnic minorities who are African-Americans, most of which are socially, economically, or educationally disadvantaged. We have delivered science-based knowledge of honey bee grooming behavior to students, beekeepers, and stakeholders through formal or informal educational programs. We created a new class as formal classroom instruction at Central State University, created new laboratory instructions on the grooming assay; developed curriculum and innovative teaching methodologies for bee biology and beekeeping; created summer research internships for CSU students. We also held online webinars and workshops to talk about honey bee queen biology with CSU extension team. Changes/Problems:We planned to hire a postdoctoral researcher to continue the project because the last postdoc researcher left the position in the winter of 2022. But the hiring process took too long. We are now finished with this project, and we did have enough time to hire a postdoc. What opportunities for training and professional development has the project provided?We provided training for CSU undergraduate students on bee research, one student learned honeybee biology and behavior, how to form a hypothesis, how to collect samples, collect data, and learned how to use the t test for statistical analysis. How have the results been disseminated to communities of interest?Presentations are local beekeeper association meetings, Ohio State Beekeeper Association meeting, CSU open house day, homecoming day, tours for the public What do you plan to do during the next reporting period to accomplish the goals?We are in the revision of one manuscript to publish our data on mite-biting behavior between Ohio feral colonies, which were our foundational colonies when we started our breeding program. We are in the middle of analyzing the data of transcriptomic of bee mandibles from high and low biting colonies.

Impacts
What was accomplished under these goals? We havepublished the results comparing mandibles between Indiana high-biting breeding stocks and commercial colonies. The paper revealedthe evolutionary morphology of mandibles among mite-resistant populations ofApis mellifera from Indiana and Ohioto elucidate the behavioral microevolution of grooming and mite-biting behavior. The paper is published athttps://doi.org/10.3389/fevo.2021.638308 We also carried results of the RNA-seq data ofmite biting behavior via transcriptomic profiles of bee mandibles. We developed a phone application nowto increase the accuracy of screening MBB bees, and improve honeybee health by genetics and breeding mite-resistant bee populations.

Publications

  • Type: Journal Articles Status: Published Year Published: 2021 Citation: Morphological changes in the mandibles accompany the defensive behavior of Indiana mite biting honey bees against Varroa destructor J Smith, XL Cleare, K Given, H Li-Byarlay, 2021, Frontiers in Ecology and Evolution, 243
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Mandibles from mite-biting bees: a multifaceted approach. M. C. Farrell, H Li-Byarlay, 2022 International Branch Virtual Symposium, Entomological Society of America.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Mandible changes and grooming behavior between two honey bee species. H Li-Byarlay, 2023 International Branch Virtual Symposium, Entomological Society of America.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Grooming behavior in two honey bee species. H Li-Byarlay, 2023 International Conference on Pollinator Biology, Health, Policy. The Penn Stater Conference Center, June 3 - 6, 2023, University Park, PA 16802.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Genetics and breeding of mite biting bees for resilience to Varroa mites. H Li-Byarlay, 2023 USDA Pollinator Health PD Meeting (virtual). 9/28/2023
  • Type: Conference Papers and Presentations Status: Published Year Published: 2024 Citation: Genes and Genomics underlying the Social Behavior of Honey Bees. H Li-Byarlay, Biology Seminar, University of Florida, Gainsville, FL. 1/9/2024
  • Type: Conference Papers and Presentations Status: Published Year Published: 2021 Citation: Potential mandible differences in the transcriptome among different bee stocks: A multifaceted approach. M. C. Farrell, H Li-Byarlay, 2021 Entomological Society of America Annual Conference (virtual)
  • Type: Conference Papers and Presentations Status: Published Year Published: 2021 Citation: Comparison of grooming behavior between two honey bee species: Apis mellilfera and Apis cerana. H. Belton, H. Li-Byarlay, 2021 Entomological Society of America Annual Conference (virtual)
  • Type: Other Status: Published Year Published: 2021 Citation: Integrative Research to Study Social Honeybees: Genes, Behavior, Brain, and Health. H. Li-Byarlay, 2021. Entomology Seminar, Michigan State University
  • Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Gene Expression Analysis and the Grooming Behavior in Different Ages of Worker Bees. H. Li-Byarlay 2023. Kentucky Queen Breeding Association Workshop


Progress 06/01/22 to 05/31/23

Outputs
Target Audience:Undergraduate students in agriculture and biology majors, beekeepers and bee farmers in Ohio and Midwest. Changes/Problems:The postdoc has changed position, so a new postdoc may need to be hired to finalize all the data analysis and finish writingthe manuscript. What opportunities for training and professional development has the project provided? Nothing Reported How have the results been disseminated to communities of interest?We have presented results in beekeeper meetings at Central Ohio Beekeeper Association meeting in Feb 2023. What do you plan to do during the next reporting period to accomplish the goals?We plan to finish writing the manuscript, and publish the results.

Impacts
What was accomplished under these goals? The RNA-seq data was analyzed, and the postdoc was working on the ms.

Publications

  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: 4/25/2022 Mandibles from mite-biting bees: a multifaceted approach ESA IB branch symposium, International Branch Entomological Society of America Virtual symposium


Progress 06/01/21 to 05/31/22

Outputs
Target Audience:Undergraduate students in CESTA Changes/Problems:Challenges: Impact of Covid 19 on student motivation to do research What opportunities for training and professional development has the project provided?Students Xaryn Cleare, Mataio Gibson, Frances Gibson participated with CSU Land-Grant researchers for experiential learning opportunities, conducted research projects with CSU bee lab. How have the results been disseminated to communities of interest? Project information and progress are posted on lab website. Program director shared information at international and national conference. CESTA Newsletters What do you plan to do during the next reporting period to accomplish the goals? Our plan is to publish the manuscript on the mandible transcriptomics, microCT scanning, and behavioral assay.

Impacts
What was accomplished under these goals? Goal 1: Compare the evolutionary morphology of mandibles among mite resistant populations of Apis mellifera and Apis cerana to elucidate the potential behavioral microevolution of mite biting behavior. We have scanned 24 mandible samples to compare the morphology between 1 day vs 8-day old workers, and high biting vs low biting workers. Goal 2) identify transcriptomic profiles and genetic markers for mite biting behavior; To assess transcriptional differences between bees classified as "high mite biting" and "low mite biting", bees were collected from the Purdue University breeding stocks, as well as feral and commercial colonies in the Central State University bee yard. In all groups, bees were collected at both 1 day old (within 24 hours of emergence) and at 8 days old, so that any changes in the transcriptomic profile throughout development could be identified. To corroborate the biting ability of these bees, mites were collected from the bottom boards at several time points throughout the season (summer and fall), and their damage was evaluated by looking at damage to the body and legs. In addition, a modified behavioral assay was conducted to look for behavioral differences between the high- and low-biting bees. RNA-seq was conducted on total RNA extracted from mandibles of high and low-biting bees. Analysis has begun. The principal component analysis showed the grouping of samples from 1 day old High-biting (HB) and low biting (LB) Purdue bees. Clustering suggests that group HB11 and HB12 are closely clustered, LB12 and LB13 are closely clustered, but a third cluster containing LB11 and HB13 is seen in the lower portion of the graph. Distinct expression profiles were observed between high biting and low biting bees at 1 day old. In addition to sequence analysis, the mandibles from the Purdue high- and low-biting bees were carefully dissected and taken to Ohio State University's Center for Electron Microscopy and Analysis (CEMAS), where they were imaged with the Heliscan MicroCT scanner. The mandibles height, width, long edge (biting edge), short edge, and ridge length are being quantified using Aviso software at CEMAS. Whole heads of both high- and low-biting bees are being stained with iodine to measure the volume of the muscles controlling mandible abduction and adduction, to see if there are any differences in muscle volume which could account for biting differences. Goal 3) apply knowledge of in-situ techniques and a phone application to increase the accuracy of screening MBB bees, and improve honeybee health by genetics and breeding mite resistant bee populations. There are a number of existing machine learning based technologies that can be used for image classification. Inspired by biological processes, A Convolutional Neuron Networks (CNN) is one of the most popular deep learning structures that require little prior knowledge or pre-processing compared to other image classification algorithms. Also, a pre-trained neural networks can be adapted to identify new objects using transfer learning and a relatively small database (Hoo-Chang Shin, 2016), which is particularly suitable for the goal of this work. The Mite Database consists of a total of 652 images and manually categorized into three classes: Healthy Mite, Damaged Mite, and Unknown. The Healthy Mite class contains 154 microscope mite images without major damages. The Damaged Mite class contains 198 images of mite with severe body damages. The Unknown class contains 300 random object images collected from flickr.com. Those random images are under the Public Domain Dedication (CC0) category, so that one can freely exploit and use them without restriction under copyright or database law. Figure A: Sample images from the Mite Database. 80% of the images in the database are randomly selected as the training data, 10% as validation date and the rest 10% as the test data. The test data is used only once for reporting the accuracy. The accuracy of the mite classification is measured by using the well-known and widely used formula as follows: where TP is the number of True Positives, TN is the number of True Negatives, FP is the number of False Positives and FN is the number of False Negatives. A number of current state-of-the-art mobile size neural networks are fine-tuned with transfer learning and the experimental results are reported in the later sections. Each of these custom neural networks can be used to develop a Phone App for mite classification using Android Studio, Google's official integrated development environment. 2. Phone app development Results Since the goal is to create a Mobile App, a number of mobile size Convolutional Neural Networks are considered in this study, which includes: ResNet50 (He, 2016) MobileNet v2 (Sandler, 2018) EfficientNet B0-B4 (Mingxing Tan, 2019) Each of these networks also has a Tensorflow Liter version that can further increase the inferencing speed for phones, tablets and embedded devices with limited computational power. A default parameter setup in Keras is used to re-train each neural network. We use 0.01 for the learning rate, 32 for the batch size, and ran 10 epochs with 16 iterations in each epoch. When changing these parameters, no significant improvement is observed in this experiment. When running the same programs multiple times, 2% to 7% accuracy difference may occur due to random split of images. The compared performance for the accuracy between tested neural networks. The experimental results suggest that these pre-trained networks are transferable to healthy/damaged mite classification without losing much of the performance, considering the fact that a clear distinction between a healthy mite and a damaged mite is sometimes difficult to tell even by the human eyes. Classification results using EfficientNet_B3. Red labelled images are misclassified. The best accuracy is achieved at 93.9% using EfficientNet, which is one of most accurate and efficient neural networks developed recently. ResNet50 achieves 80.3% accuracy and MobileNet v2 achieves 87.9% accuracy, while accuracy among EfficientNet B0-B4 vary from 83.3% to 93.9%. A higher accuracy might be achieved via a more complex neural network with the sacrifice of speed. It is safe to say the experimental result proves the concept that healthy mites and damaged mites can be distinguished using Artificial Intelligence, and this technology can be used to assist bee farmers to count the damaged mite from each colony using an average smart phone. When a precise count is not necessary to compare different colonies. an approximately 90% classification accuracy is realistically acceptable. Finally, an Android phone app was made based on EfficientNet_lite3 which has the highest accuracy. We are able to develop an Android App "Mite Classification" with sample classification results. In addition to studying the mandibles of the mite-biting bees, an app was developed to take microscopy images of mites, and assess whether or not they have been bitten. This app has been created for android, and is being made available for download. We have tested the app with known mite images, and the app is reporting data accurately based on test images. Dr. Hongmei Li-Byarlay conducted a training with the agricultural extension office to train beekeepers how to use microscopes to analyze the mite-biting percentage of their own colonies, to breed for higher biting bees. Our project is to improve the resilience trait of European Honey Bees against Varroa mites in Ohio and in Midwest. We provided newest solution to mitigate the colony losses in Ohio and other Midwest regions. As the newest 1890 land-grant institution in collaborating with Purdue University, we enhanced CSU's research capacity in pollinator health and sustainable apiculture. We have trained 3 minority students (Xaryn Cleare, Mataio Gibson, Frances Gibson, ) to join the future USA's sustainable agriculture work force.

Publications


    Progress 06/01/20 to 05/31/21

    Outputs
    Target Audience:Target audiencesare undergraduate students at Central State University, individuals, and groups of beekeepers and communitiesin Ohio. Students at Central State University are racial and ethnic minorities who are African-Americans, most of whichare socially, economically, or educationally disadvantaged. We havedelivered science-based knowledge of honey bee grooming behavior to students, beekeepers, and stakeholdersthrough formal or informal educational programs. We created a new class as formal classroom instruction at Central State University, created new laboratory instructions on the grooming assay; developedcurriculum andinnovative teaching methodologies for bee biology and beekeeping; created summer research internships for CSU students. We also held online webinars and workshops to talk about honey bee queen biology with CSUextension team. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?Two Central State students: Mr. Jawwaad Sabree and Mr. Lee Burrows were hired and trained for the website development. Both of them are Computer Science majors. We had weekly Zoom meetings with Purdue university and our students to discuss their work progresses. Together we are building a website that can collect mite images from beekeepers to build a healthy / damaged mite database. During this study, our students applied their knowledge learned from our Computer Science courses, such as CPS2215 - Internet Web Essentials, and honed their skills using HTML, Java Script, and Cascading Style Sheets (CSS). Working with the front and back end of the website, they also learned new knowledge and techniques that were not covered in the classroom, such as PHP programming, Bootstrap framework, Github setting, and honey bee ecology. This project helped our students tremendously. They are exposed to cutting-edge technologies and improved their programming, collaboration, communication, and presentation skills. They gained valuable experiences and were inspired and intrigued to learn more. Mr. Burrows and I started separate meetings for intact/damaged mite classification using machine learning platforms such as Tensorflow. We covered machine learning basics and had test runs on Google Colaboratory and Ubuntu OS. Mr. Burrows completed his senior project based on what he learned in this side-project. We were also glad to know that Mr. Sabree was admitted by Georgia State University as an undergraduate student, and Mr. Burrows is now considered by Bowling Green University and Wright State University as a graduate student. Wright State University already offered a full assistantship to Mr. Burrows. How have the results been disseminated to communities of interest? Date Title Meeting/Institution Funding presentater audience Invited hosted by 120 4/12/2021 What do we learn about Genes, DNA methylation, Brain, and Social Behavior of Honey Bees University College of London online invited HML 100 112 2/11/2021 My Experience at an HBCU and diversity /inclusion /equity in the community of Entomology University of Kentucky online invited HML 60 108 1/26/2021 Grooming and biting behavior of honeybees Global Animal Behaviour Twitter Conference online free HML 30 107 1/7/2021 The mite biting and grooming behavior of honeybees ABRC online LGF/LGS HML 180 104 11/11-20, 2020 Analysis of honeybee queen quality (Apis mellifera) between feral and commercial bess ESA online 8507 XLC 100 102 11/7/2020 Mite Biting Behavior of Honeybees and the potential mechanisms Ohio State Beekeeper Association online invited HML 1000 https://youtu.be/LiBaeybBNiA 100 10/2/2020 Distilling 10 years of research into one hour: Gene, Genome, Behavior, and Physiology of Honey Bee Department Seminar PSU virtual - HML 50 invited What do you plan to do during the next reporting period to accomplish the goals?1) Finish writing and publish the manuscript of comparingthe evolutionary morphology of mandibles among mite resistant populations ofApis melliferaandApisceranato elucidate the potential behavioral microevolution of mite biting behavior 2) carry out the RNAseq experiments to identify genetic markers based on transcriptomic profiles of bee mandibles, and 3) develop a phone application for beekeepers to screen forMite-biterbees, and improve honeybee health by genetics and breeding mite-resistant bee populations.

    Impacts
    What was accomplished under these goals? 1)We havecompared the evolutionary morphology of mandibles between mite-biters of Indiana and commercial honeybees. A new peer-reviewed paper has been published in the journal of Frontiers in Ecology and Evolution. 2) We havecompared the evolutionary morphology of mandibles among resistant populations ofApis melliferaandApisceranato elucidate the potential behavioral microevolution of mite biting behavior. A published is in preparation and will be submitted soon. 3) We finish the hiring process and preparation for a new postdoc researcher who will investigate the genetic basis of mite biting behavior via transcriptomic profiles of bee mandibles. 4) We have worked with Purdue University to create a website (https://dev.beekeeper.ceris.purdue.edu/home) to help beekeepers to upload mite pictures, and provide reports for mite biting behavior, which willimprove honeybee health by genetics and breeding mite-resistant bee populations.

    Publications

    • Type: Journal Articles Status: Published Year Published: 2021 Citation: Smith J, Cleare X, Given K, Li-Byarlay H, 2021. Morphological changes in the mandibles accompany the defensive behavior of Indiana mite biting honey bees against Varroa destructor, Frontiers in Ecology and Evolution, doi: 10.3389/fevo.2021.638308
    • Type: Websites Status: Under Review Year Published: 2021 Citation: https://dev.beekeeper.ceris.purdue.edu/home
    • Type: Websites Status: Published Year Published: 2020 Citation: https://sites.google.com/view/libyarlaybee/Home/bee-health-genetics-and-breeding?authuser=0
    • Type: Conference Papers and Presentations Status: Published Year Published: 2020 Citation: The mite biting and grooming behavior of honeybees
    • Type: Conference Papers and Presentations Status: Published Year Published: 2021 Citation: 1/26/2021 Grooming and biting behavior of honeybees Global Animal Behaviour Twitter Conference online
    • Type: Other Status: Published Year Published: 2021 Citation: 2/11/2021 My Experience at an HBCU and diversity /inclusion /equity in the community of Entomology University of Kentucky online
    • Type: Other Status: Published Year Published: 2021 Citation: 4/12/2021, What do we learn about Genes, DNA methylation, Brain, and Social Behavior of Honey Bees, University College of London, dept seminar online