Artificial intelligence can revolutionize scientific projects, but scientists face challenges in reusing, integrating, and deploying cost-effective and high-quality machine learning solutions. Determining suitable algorithms and parameters is difficult, especially for non-programmer scientists. Some algorithms, like deep learning-based methods, offer flexibility but require extensive training on annotated data. This poses a hurdle in labor-intensive tasks like biological image segmentation that relies on expert annotations. In this paper, we present a data-driven framework designed to assist scientists in selecting, reusing, and training machine learning solutions for microscopy image segmentation. The framework is based on establishing a mapping between object morphology features and the optimal segmentation algorithms and settings for individual objects. This mapping is iteratively refined through a combination of unsupervised learning and active learning iterations. To expedite convergence, objects are initially clustered based on their morphology. In each active learning iteration, the most informative and uncertain samples are selected and queried within a specific cluster. Through a biological case study, we demonstrate that our method enables the selection and training of segmentation algorithms specific to object types. Additionally, the selective requests for user input significantly reduce the number of user interactions required for this task.

Machine learning techniques have revolutionised scientific software projects. Scientists are continuously looking for novel approaches to production-quality reuse of machine learning solutions and to make them available to other components of the project with satisfactory quality and low costs. However, scientists often have limited knowledge about how to effectively reuse and adjust machine learning solutions in their particular scientific project. One challenge is that many machine learning solutions require parameter tuning based on the input data to achieve satisfactory results, which is difficult and cumbersome for users not familiar with machine learning. Autotuning is the common technique for potentially adjusting the parameters based on the data, but it requires a well-defined objective function to optimize for. Such an objective function is commonly unknown in exploratory scientific research such as biological image segmentation tasks. In this paper, we propose a framework based on the novel combination of autotuning and active learning to ease and partially automate the reuse effort of machine learning solutions for scientists in biological image segmentation cases. Underlying this combination is a mapping between an object type and specific parameters applied during the segmentation process. This mapping is iteratively adjusted by asking users for visual feedback. We then through a biological case study demonstrate that our method enables tuning of the segmentation specifically to object types, while the selective requests of user input reduce the number of user interactions required for this task.

The genome is packed into the cell nucleus in the form of chromatin. Biochemical approaches have revealed that chromatin is packed within domains, which group into larger domains, and so forth. Such hierarchical packing is equally visible in super-resolution microscopy images of large-scale chromatin organization. While previous work has suggested that chromatin is partitioned into distinct domains via microphase separation, it is unclear how these domains organize into this hierarchical packing. A particular challenge is to find an image analysis approach that fully incorporates such hierarchical packing, so that hypothetical governing mechanisms of euchromatin packing can be compared against the results of such an analysis. Here, we obtain 3D STED super-resolution images from pluripotent zebrafish embryos labeled with improved DNA fluorescence stains, and demonstrate how the hierarchical packing of euchromatin in these images can be described as multiplicative cascades. Multiplicative cascades are an established theoretical concept to describe the placement of ever-smaller structures within bigger structures. Importantly, these cascades can generate artificial image data by applying a single rule again and again, and can be fully specified using only four parameters. Here, we show how the typical patterns of euchromatin organization are reflected in the values of these four parameters. Specifically, we can pinpoint the values required to mimic a microphase-separated state of euchromatin. We suggest that the concept of multiplicative cascades can also be applied to images of other types of chromatin. Here, cascade parameters could serve as test quantities to assess whether microphase separation or other theoretical models accurately reproduce the hierarchical packing of chromatin.

The proteins involved in smooth muscle's molecular contractile mechanism - the anti-parallel motion of actin and myosin filaments driven by myosin heads interacting with actin - are found as different isoforms. While their expression levels are altered in disease states, their relevance to the mechanical interaction of myosin with actin is not sufficiently understood. Here, we analyzed in vitro actin filament propulsion by smooth muscle myosin for [Formula: see text]-actin ([Formula: see text]A), [Formula: see text]-actin-tropomyosin-[Formula: see text] ([Formula: see text]A-Tm[Formula: see text]), [Formula: see text]-actin-tropomyosin-[Formula: see text] ([Formula: see text]A-Tm[Formula: see text]), [Formula: see text]-actin ([Formula: see text]A), [Formula: see text]-actin-tropomyosin-[Formula: see text] ([Formula: see text]A-Tm[Formula: see text]), and [Formula: see text]-actin-tropomoysin-[Formula: see text] ([Formula: see text]A-Tm[Formula: see text]). Actin sliding analysis with our specifically developed video analysis software followed by statistical assessment (Bootstrapped Principal Component Analysis) indicated that the in vitro motility of [Formula: see text]A, [Formula: see text]A, and [Formula: see text]A-Tm[Formula: see text] is not distinguishable. Compared to these three 'baseline conditions', statistically significant differences ([Formula: see text]) were: [Formula: see text]A-Tm[Formula: see text] - actin sliding velocity increased 1.12-fold, [Formula: see text]A-Tm[Formula: see text] - motile fraction decreased to 0.96-fold, stop time elevated 1.6-fold, [Formula: see text]A-Tm[Formula: see text] - run time elevated 1.7-fold. We constructed a mathematical model, simulated actin sliding data, and adjusted the kinetic parameters so as to mimic the experimentally observed differences: [Formula: see text]A-Tm[Formula: see text] - myosin binding to actin, the main, and the secondary myosin power stroke are accelerated, [Formula: see text]A-Tm[Formula: see text] - mechanical coupling between myosins is stronger, [Formula: see text]A-Tm[Formula: see text] - the secondary power stroke is decelerated and mechanical coupling between myosins is weaker. In summary, our results explain the different regulatory effects that specific combinations of actin and smooth muscle tropomyosin have on smooth muscle actin-myosin interaction kinetics.