Lingfeng Duan

Combining high-throughput phenotyping and genome-wide association studies to reveal natural genetic variation in rice

Even as the study of plant genomics rapidly develops through the use of high-throughput sequencing techniques, traditional plant phenotyping lags far behind. Here we develop a high-throughput rice phenotyping facility (HRPF) to monitor 13 traditional agronomic traits and 2 newly defined traits during the rice growth period. Using genome-wide association studies (GWAS) of the 15 traits, we identify 141 associated loci, 25 of which contain known genes such as the Green Revolution semi-dwarf gene, SD1. Based on a performance evaluation of the HRPF and GWAS results, we demonstrate that high-throughput phenotyping has the potential to replace traditional phenotyping techniques and can provide valuable gene identification information. The combination of the multifunctional phenotyping tools HRPF and GWAS provides deep insights into the genetic architecture of important traits.

Plant phenomics and high-throughput phenotyping accelerating rice functional genomics using multidisciplinary technologies

The functional analysis of the rice genome has entered into a high-throughput stage, and a project named RICE2020 has been proposed to determine the function of every gene in the rice genome by the year 2020. However, as compared with the robustness of genetic techniques, the evaluation of rice phenotypic traits is still performed manually, and the process is subjective, inefficient, destructive and error-prone. To overcome these limitations and help rice phenomics more closely parallel rice genomics, reliable, automatic, multifunctional, and high-throughput phenotyping platforms should be developed. In this article, we discuss the key plant phenotyping technologies, particularly photonics-based technologies, and then introduce their current applications in rice (wheat or barley) phenomics. We also note the major challenges in rice phenomics and are confident that these reliable high-throughput phenotyping tools will give plant scientists new perspectives on the information encoded in the rice genome.

A novel machine-vision-based facility for the automatic evaluation of yield-related traits in rice

The evaluation of yield-related traits is an essential step in rice breeding, genetic research and functional genomics research. A new, automatic, and labor-free facility to automatically thresh rice panicles, evaluate rice yield traits, and subsequently pack filled spikelets is presented in this paper. Tests showed that the facility was capable of evaluating yield-related traits with a mean absolute percentage error of less than 5% and an efficiency of 1440 plants per continuous 24 h workday.

High-throughput measurement of rice tillers using a conveyor equipped with x-ray computed tomography

Tillering is one of the most important agronomic traits because the number of shoots per plant determines panicle number, a key component of grain yield. The conventional method of counting tillers is still manual. Under the condition of mass measurement, the accuracy and efficiency could be gradually degraded along with fatigue of experienced staff. Thus, manual measurement, including counting and recording, is not only time consuming but also lack objectivity. To automate this process, we developed a high-throughput facility, dubbed high-throughput system for measuring automatically rice tillers (H-SMART), for measuring rice tillers based on a conventional x-ray computed tomography (CT) system and industrial conveyor. Each pot-grown rice plant was delivered into the CT system for scanning via the conveyor equipment. A filtered back-projection algorithm was used to reconstruct the transverse section image of the rice culms. The number of tillers was then automatically extracted by image segmentation. To evaluate the accuracy of this system, three batches of rice at different growth stages (tillering, heading, or filling) were tested, yielding absolute mean absolute errors of 0.22, 0.36, and 0.36, respectively. Subsequently, the complete machine was used under industry conditions to estimate its efficiency, which was 4320 pots per continuous 24 h workday. Thus, the H-SMART could determine the number of tillers of pot-grown rice plants, providing three advantages over the manual tillering method: absence of human disturbance, automation, and high throughput. This facility expands the application of agricultural photonics in plant phenomics.

High-throughput phenotyping and QTL mapping reveals the genetic architecture of maize plant growth

Combining high-throughput phenotyping and large-scale QTL mapping dissects the dynamic genetic architecture of maize development by using a RIL population. With increasing demand for novel traits in crop breeding, the plant research community faces the challenge of quantitatively analyzing the structure and function of large numbers of plants. A clear goal of high-throughput phenotyping is to bridge the gap between genomics and phenomics. In this study, we quantified 106 traits from a maize (Zea mays) recombinant inbred line population (n = 167) across 16 developmental stages using the automatic phenotyping platform. Quantitative trait locus (QTL) mapping with a high-density genetic linkage map, including 2,496 recombinant bins, was used to uncover the genetic basis of these complex agronomic traits, and 988 QTLs have been identified for all investigated traits, including three QTL hotspots. Biomass accumulation and final yield were predicted using a combination of dissected traits in the early growth stage. These results reveal the dynamic genetic architecture of maize plant growth and enhance ideotype-based maize breeding and prediction.

Panicle-SEG: a robust image segmentation method for rice panicles in the field based on deep learning and superpixel optimization

BackgroundRice panicle phenotyping is important in rice breeding, and rice panicle segmentation is the first and key step for image-based panicle phenotyping. Because of the challenge of illumination differentials, panicle shape deformations, rice accession variations, different reproductive stages and the field’s complex background, rice panicle segmentation in the field is a very large challenge.ResultsIn this paper, we propose a rice panicle segmentation algorithm called Panicle-SEG, which is based on simple linear iterative clustering superpixel regions generation, convolutional neural network classification and entropy rate superpixel optimization. To build the Panicle-SEG-CNN model and test the segmentation effects, 684 training images and 48 testing images were randomly selected, respectively. Six indicators, including Qseg, Sr, SSIM, Precision, Recall and F-measure, are employed to evaluate the segmentation effects, and the average segmentation results for the 48 testing samples are 0.626, 0.730, 0.891, 0.821, 0.730, and 76.73%, respectively. Compared with other segmentation approaches, including HSeg, i2 hysteresis thresholding and jointSeg, the proposed Panicle-SEG algorithm has better performance on segmentation accuracy. Meanwhile, the executing speed is also improved when combined with multithreading and CUDA parallel acceleration. Moreover, Panicle-SEG was demonstrated to be a robust segmentation algorithm, which can be expanded for different rice accessions, different field environments, different camera angles, different reproductive stages, and indoor rice images. The testing dataset and segmentation software are available online.ConclusionsIn conclusion, the results demonstrate that Panicle-SEG is a robust method for panicle segmentation, and it creates a new opportunity for nondestructive yield estimation.

Development of a whole-feeding and automatic rice thresher for single plant ☆

Abstract Threshing is an essential pretreatment in rice yield-related traits evaluation and rice thresher is an important study of agricultural machinery automation. However, traditional threshers aim at threshing filled grains with simple function, which are inapplicable to high-precision and automatic yield-related traits evaluation of rice. And the conventional threshing method for the traits evaluation is still manual. To improve it, we developed a whole-feeding and automatic rice thresher for single plant. The thresher adopted a hierarchical multi-roller rolling method to thresh filled and unfilled grains respectively. A fish scale sieve plate was designed to separate the grains from the straw and the whole system was controlled by Programmable Logic Controller (PLC) automatically. To evaluate the threshing system, two batches of rice plants were tested, and the results showed that the thresher had the advantages of high precision, low breakage and no residue. Thus, the thresher provides strong support for high-precision and automatic yield-related traits evaluation of rice.

A nondestructive method for estimating the total green leaf area of individual rice plants using multi-angle color images

Total green leaf area (GLA) is an important trait for agronomic studies. However, existing methods for estimating the GLA of individual rice plants are destructive and labor-intensive. A nondestructive method for estimating the total GLA of individual rice plants based on multi-angle color images is presented. Using projected areas of the plant in images, linear, quadratic, exponential and power regression models for estimating total GLA were evaluated. Tests demonstrated that the side-view projected area had a stronger relationship with the actual total leaf area than the top-projected area. And power models fit better than other models. In addition, the use of multiple side-view images was an efficient method for reducing the estimation error. The inclusion of the top-view projected area as a second predictor provided only a slight improvement of the total leaf area estimation. When the projected areas from multi-angle images were used, the estimated leaf area (ELA) using the power model and the actual leaf area had a high correlation coefficient (R2 > 0.98), and the mean absolute percentage error (MAPE) was about 6%. The method was capable of estimating the total leaf area in a nondestructive, accurate and efficient manner, and it may be used for monitoring rice plant growth.

High-Throughput Estimation of Yield for Individual Rice Plant Using Multi-angle RGB Imaging

Modern breeding technologies are capable of producing hundreds of new varieties daily, so fast, simple and effective methods for screening valuable candidate plant materials are urgently needed. Final yield is a significant agricultural trait in rice breeding. In the screening and evaluation of the rice varieties, measuring and evaluating rice yield is essential. Conventional means of measuring rice yield mainly depend on manual determination, which is tedious, labor-intensive, subjective and error-prone, especially when large-scale plants were to be investigated. This paper presented an in vivo, automatic and high-throughput method to estimate the yield of individual pot-grown rice plant using multi-angle RGB imaging and image analysis. In this work, we demonstrated a new idea of estimating rice yield from projected panicle area, projected area of leaf and stem and fractal dimension. 5-fold cross validation showed that the predictive error was 7.45%. The constructed model achieved promising results on rice plants grown both in-door and out-door. The presented work has the potential of accelerating yield estimation and would be a promising impetus for plant phenomics.

Novel Digital Features Discriminate Between Drought Resistant and Drought Sensitive Rice Under Controlled and Field Conditions

Dynamic quantification of drought response is a key issue both for variety selection and for functional genetic study of rice drought resistance. Traditional assessment of drought resistance traits, such as stay-green and leaf-rolling, has utilized manual measurements, that are often subjective, error-prone, poorly quantified and time consuming. To relieve this phenotyping bottleneck, we demonstrate a feasible, robust and non-destructive method that dynamically quantifies response to drought, under both controlled and field conditions. Firstly, RGB images of individual rice plants at different growth points were analyzed to derive 4 features that were influenced by imposition of drought. These include a feature related to the ability to stay green, which we termed greenness plant area ratio (GPAR) and 3 shape descriptors [total plant area/bounding rectangle area ratio (TBR), perimeter area ratio (PAR) and total plant area/convex hull area ratio (TCR)]. Experiments showed that these 4 features were capable of discriminating reliably between drought resistant and drought sensitive accessions, and dynamically quantifying the drought response under controlled conditions across time (at either daily or half hourly time intervals). We compared the 3 shape descriptors and concluded that PAR was more robust and sensitive to leaf-rolling than the other shape descriptors. In addition, PAR and GPAR proved to be effective in quantification of drought response in the field. Moreover, the values obtained in field experiments using the collection of rice varieties were correlated with those derived from pot-based experiments. The general applicability of the algorithms is demonstrated by their ability to probe archival Miscanthus data previously collected on an independent platform. In conclusion, this image-based technology is robust providing a platform-independent tool for quantifying drought response that should be of general utility for breeding and functional genomics in future.