Self-supervised maize kernel classification and segmentation for embryo identification.

Introduction: Computer vision and deep learning (DL) techniques have succeeded in a wide range of diverse fields. Recently, these techniques have been successfully deployed in plant science applications to address food security, productivity, and environmental sustainability problems for a growing global population. However, training these DL models often necessitates the large-scale manual annotation of data which frequently becomes a tedious and time-and-resource- intensive process. Recent advances in self-supervised learning (SSL) methods have proven instrumental in overcoming these obstacles, using purely unlabeled datasets to pre-train DL models. Methods: Here, we implement the popular self-supervised contrastive learning methods of NNCLR Nearest neighbor Contrastive Learning of visual Representations) and SimCLR (Simple framework for Contrastive Learning of visual Representations) for the classification of spatial orientation and segmentation of embryos of maize kernels. Maize kernels are imaged using a commercial high-throughput imaging system. This image data is often used in multiple downstream applications across both production and breeding applications, for instance, sorting for oil content based on segmenting and quantifying the scutellum's size and for classifying haploid and diploid kernels. Results and discussion: We show that in both classification and segmentation problems, SSL techniques outperform their purely supervised transfer learning-based counterparts and are significantly more annotation efficient. Additionally, we show that a single SSL pre-trained model can be efficiently finetuned for both classification and segmentation, indicating good transferability across multiple downstream applications. Segmentation models with SSL-pretrained backbones produce DICE similarity coefficients of 0.81, higher than the 0.78 and 0.73 of those with ImageNet-pretrained and randomly initialized backbones, respectively. We observe that finetuning classification and segmentation models on as little as 1% annotation produces competitive results. These results show SSL provides a meaningful step forward in data efficiency with agricultural deep learning and computer vision.

Acoustofluidic Droplet Sorter Based on Single Phase Focused Transducers.

Droplet microfluidics has revolutionized the biomedical and drug development fields by allowing for independent microenvironments to conduct drug screening at the single cell level. However, current microfluidic sorting devices suffer from drawbacks such as high voltage requirements (e.g., >200 Vpp), low biocompatibility, and/or low throughput. In this article, a single-phase focused transducer (SPFT)-based acoustofluidic chip is introduced, which outperforms many microfluidic droplet sorting devices through high energy transmission efficiency, high accuracy, and high biocompatibility. The SPFT-based sorter can be driven with an input power lower than 20 Vpp and maintain a postsorting cell viability of 93.5%. The SPFT sorter can achieve a throughput over 1000 events per second and a sorting purity up to 99.2%. The SPFT sorter is utilized here for the screening of doxorubicin cytotoxicity on cancer and noncancer cells, proving its drug screening capability. Overall, the SPFT droplet sorting device shows great potential for fast, precise, and biocompatible drug screening.

Recent advances in droplet microfluidics for enzyme and cell factory engineering.

Enzymes and cell factories play essential roles in industrial biotechnology for the production of chemicals and fuels. The properties of natural enzymes and cells often cannot meet the requirements of different industrial processes in terms of cost-effectiveness and high durability. To rapidly improve their properties and performances, laboratory evolution equipped with high-throughput screening methods and facilities is commonly used to tailor the desired properties of enzymes and cell factories, addressing the challenges of achieving high titer and the yield of the target products at high/low temperatures or extreme pH, in unnatural environments or in the presence of unconventional media. Droplet microfluidic screening (DMFS) systems have demonstrated great potential for exploring vast genetic diversity in a high-throughput manner (>106/h) for laboratory evolution and have been increasingly used in recent years, contributing to the identification of extraordinary mutants. This review highlights the recent advances in concepts and methods of DMFS for library screening, including the key factors in droplet generation and manipulation, signal sources for sensitive detection and sorting, and a comprehensive summary of success stories of DMFS implementation for engineering enzymes and cell factories during the past decade.

A negative dielectrophoresis and gravity-driven flow-based high-throughput and high-efficiency cell-sorting system.

We present a negative dielectrophoresis (n-DEP)-based cell separation system for high-throughput and high-efficiency cell separation. To achieve a high throughput, the proposed system comprises macro-sized channel and cantilever-type electrode (CE) arrays (L × W × H = 150 µm × 500 µm × 50 µm) to generate n-DEP force. For high efficiency, double separation modules, which have macro-sized channels and CE arrays in each separation module, are employed. In addition, flow regulators to precisely control the hydrodynamic force are allocated for each outlet. Because the hydrodynamic force and the n-DEP force acting on the target cell are the main determinants of the separation efficiency, we evaluate the theoretical amount of hydrodynamic force and n-DEP force acting on each target cell. Based on theoretical results, separation conditions are experimentally investigated. Finally, to demonstrate the separation performance, we performed the separation of target cells (live K562) from nontarget cells (dead K562) under conditions of low voltage (7Vp-p with 100 kHz) and a flow rate of 15 µL•min⁻¹, 6 µL•min⁻¹, and 8 µL•min⁻¹ in outlets 1, 2, and 3, respectively. The system can separate target cells with 95% separation efficiency in the case of the ratio of 5:1 (live K562:dead K562).