In gastric cancer (GC), there are four molecular subclasses that indicate whether patients respond to chemotherapy or immunotherapy, according to the TCGA. In clinical practice, however, not every patient undergoes molecular testing. Many laboratories have used well-implemented in situ techniques (IHC and EBER-ISH) to determine the subclasses in their cohorts. Although multiple stains are used, we show that a staining approach is unable to correctly discriminate all subclasses. As an alternative, we trained an ensemble convolutional neuronal network using bagging that can predict the molecular subclass directly from hematoxylin-eosin histology. We also identified patients with predicted intra-tumoral heterogeneity or with features from multiple subclasses, which challenges the postulated TCGA-based decision tree for GC subtyping. In the future, deep learning may enable targeted testing for molecular subtypes and targeted therapy for a broader group of GC patients.

The mechanics of fibronectin-rich extracellular matrix regulate cell physiology in a number of diseases, prompting efforts to elucidate cell mechanosensing mechanisms at the molecular and cellular scale. Here, the use of fibronectin-functionalized silicone elastomers that exhibit considerable frequency-dependence in viscoelastic properties unveiled the presence of two cellular processes that respond discreetly to substrate mechanical properties.Soft elastomers supported efficient focal adhesion maturation and fibroblast spreading due to an apparent stiff surface layer. However, soft elastomers did not enable cytoskeletal and fibroblast polarization; elastomers with high cross-linking and low deformability were required for polarization. The underlying reason for this behavior was the inability of soft elastomeric substrates to resist traction forces, rather than a lack of sufficient traction force generation; accordingly, mild inhibition of actomyosin contractility rescued fibroblast polarization even on the softer elastomers. Our findings help reconcile previously proposed local and global models of cell mechanosensing by demonstrating the differential dependence of substrate mechanics on distinct cellular processes.

There is a lot of recent interest in the field of computational pathology, as many algorithms are introduced to detect, for example, cancer lesions or molecular features. However, there is a large gap between artificial intelligence (AI) technology and practice, since only a small fraction of the applications is used in routine diagnostics. The main problems are the transferability of convolutional neural network (CNN) models to data from other sources and the identification of uncertain predictions. The role of tissue quality itself is also largely unknown. Here, we demonstrated that samples of the TCGA ovarian cancer (TCGA-OV) dataset from different tissue sources have different quality characteristics and that CNN performance is linked to this property. CNNs performed best on high-quality data. Quality control tools were partially able to identify low-quality tiles, but their use did not increase the performance of the trained CNNs. Furthermore, we trained NoisyEnsembles by introducing label noise during training. These NoisyEnsembles could improve CNN performance for low-quality, unknown datasets. Moreover, the performance increases as the ensemble become more consistent, suggesting that incorrect predictions could be discarded efficiently to avoid wrong diagnostic decisions.

Axonal degeneration (AxD) is a pathological hallmark of many neurodegenerative diseases. Deciphering the morphological patterns of AxD will help to understand the underlying mechanisms and develop effective therapies. Here, we evaluated the progression of AxD in cortical neurons using a novel microfluidic device together with a deep learning tool that we developed for the enhanced-throughput analysis of AxD on microscopic images. The trained convolutional neural network (CNN) sensitively and specifically segmented the features of AxD including axons, axonal swellings, and axonal fragments. Its performance exceeded that of the human evaluators. In an in vitro model of AxD in hemorrhagic stroke induced by the hemolysis product hemin, we detected a time-dependent degeneration of axons leading to a decrease in axon area, while axonal swelling and fragment areas increased. Axonal swellings preceded axon fragmentation, suggesting that swellings may be reliable predictors of AxD. Using a recurrent neural network (RNN), we identified four morphological patterns of AxD (granular, retraction, swelling, and transport degeneration). These findings indicate a morphological heterogeneity of AxD in hemorrhagic stroke. Our EntireAxon platform enables the systematic analysis of axons and AxD in time-lapse microscopy and unravels a so-far unknown intricacy in which AxD can occur in a disease context.