Abstract:Cellular force transmission across a hierarchy of molecular switchers is central to mechanobiological responses. However, current cellular force microscopies suffer from low throughput and resolution. Here we introduce and train a generative adversarial network (GAN) to paint out traction force maps of cell monolayers with high fidelity to the experimental traction force microscopy (TFM). The GAN analyzes traction force maps as an image-to-image translation problem, where its generative and discriminative neural networks are simultaneously cross-trained by hybrid experimental and numerical datasets. In addition to capturing the colony-size and substrate-stiffness dependent traction force maps, the trained GAN predicts asymmetric traction force patterns for multicellular monolayers seeding on substrates with stiffness gradient, implicating collective durotaxis. Further, the neural network can extract experimentally inaccessible, the hidden relationship between substrate stiffness and cell contractility, which underlies cellular mechanotransduction. Trained solely on datasets for epithelial cells, the GAN can be extrapolated to other contractile cell types using only a single scaling factor. The digital TFM serves as a high-throughput tool for mapping out cellular forces of cell monolayers and paves the way toward data-driven discoveries in cell mechanobiology.

What problem does this paper attempt to address?

The main goal of this paper is to address the technical challenges of traction force measurement in cell mechanics research, particularly by improving the throughput and resolution of Traction Force Microscopy (TFM). Specifically, the paper addresses these issues through the following points: 1. **Introduction of Generative Adversarial Networks (GANs)**: To overcome the limitations of traditional TFM methods in terms of throughput and resolution, the authors developed a GAN-based approach to predict the traction force maps of cell monolayers. This method can faithfully simulate experimental TFM results. 2. **Addressing Data Scarcity**: Due to the limited data provided by experimental TFM, the authors combined experimental data with numerical simulation data to train the GAN. This included both experimental data and high-throughput continuous model-generated data. 3. **Predicting Traction Force Distribution under Different Conditions**: The trained GAN can predict traction force maps for cell colonies of different sizes and under different substrate stiffness conditions. Additionally, it can predict asymmetric traction force patterns when cells are planted on substrates with stiffness gradients, indicating collective durotaxis. 4. **Revealing Hidden Relationships**: The neural network can also reveal relationships that are difficult to obtain experimentally, such as the implicit link between substrate stiffness and cell contractility, which is crucial for understanding cellular mechanotransduction. 5. **Cross-Cell Type Application**: Although the GAN was trained using data from epithelial cells, it can be applied to other types of contractile cells through simple scaling factors, demonstrating the model's generalization capability. In summary, this paper proposes a novel digital TFM (DTFM) method that significantly enhances the efficiency and applicability of traction force measurement using machine learning techniques, providing a powerful tool for research in the field of cell mechanics.

Machine learning traction force maps of cell monolayers

Enhancing robustness, precision, and speed of traction force microscopy with machine learning

Wrinkle force microscopy: a new machine learning based approach to predict cell mechanics from images

Machine learning interpretable models of cell mechanics from protein images

Cellogram: On-the-Fly Traction Force Microscopy

Traction force microscopy with optimized regularization and automated Bayesian parameter selection for comparing cells

The Continuous Evolution of 2D Cell-Traction Forces Quantification Technology

High-resolution assessment of multidimensional cellular mechanics using label-free refractive-index traction force microscopy

Generation and application of three-dimensional cell traction force

Enhanced Cell Tracking Using A GAN-based Super-Resolution Video-to-Video Time-Lapse Microscopy Generative Model

A Detailed Protocol for Cell Force Measurement by Traction Force Microscopy

Two-Layer Elastographic 3-D Traction Force Microscopy

Improved-Throughput Traction Microscopy Based on Fluorescence Micropattern for Manual Microscopy

Deep Learning Reveals How Cells Pull, Buckle, and Navigate Tissue-Like Environments

3D micropatterned traction force microscopy: a technique to control three-dimensional cell shape while measuring cell-substrate force transmission

3D Traction Force Microscopy in Biological Gels: From Single Cells to Multicellular Spheroids

Quantifying Cell Traction Forces at the Individual Fiber Scale in 3D: A Novel Approach Based on Deformable Photopolymerized Fiber Arrays

Computational Tension Mapping of Adherent Cells Based on Actin Imaging

Nanomechanical Force Mapping of Restricted Cell-To-Cell Collisions Oscillating between Contraction and Relaxation.

Two-dimensional TIRF-SIM–traction Force Microscopy (2D TIRF-SIM-TFM)

Traction Force Microscopy with DNA FluoroCubes