Enhancing Spatial Transcriptomics Analysis by Integrating Image-Aware Deep Learning Methods.

Spatial transcriptomics (ST) represents a pivotal advancement in biomedical research, enabling the transcriptional profiling of cells within their morphological context and providing a pivotal tool for understanding spatial heterogeneity in cancer tissues. However, current analytical approaches, akin to single-cell analysis, largely depend on gene expression, underutilizing the rich morphological information inherent in the tissue. We present a novel method integrating spatial transcriptomics and histopathological image data to better capture biologically meaningful patterns in patient data, focusing on aggressive cancer types such as glioblastoma and triple-negative breast cancer. We used a ResNet-based deep learning model to extract key morphological features from high-resolution whole-slide histology images. Spot-level PCA-reduced vectors of both the ResNet-50 analysis of the histological image and the spatial gene expression data were used in Louvain clustering to enable image-aware feature discovery. Assessment of features from image-aware clustering successfully pinpointed key biological features identified by manual histopathology, such as for regions of fibrosis and necrosis, as well as improved edge definition in EGFR-rich areas. Importantly, our combinatorial approach revealed crucial characteristics seen in histopathology that gene-expression-only analysis had missed.Supplemental Material: https://github.com/davcraig75/song_psb2014/blob/main/SupplementaryData.pdf.

Fluorescently labeled nuclear morphology is highly informative of neurotoxicity.

Neurotoxicity can be detected in live microscopy by morphological changes such as retraction of neurites, fragmentation, blebbing of the neuronal soma and ultimately the disappearance of fluorescently labeled neurons. However, quantification of these features is often difficult, low-throughput, and imprecise due to the overreliance on human curation. Recently, we showed that convolutional neural network (CNN) models can outperform human curators in the assessment of neuronal death from images of fluorescently labeled neurons, suggesting that there is information within the images that indicates toxicity but that is not apparent to the human eye. In particular, the CNN's decision strategy indicated that information within the nuclear region was essential for its superhuman performance. Here, we systematically tested this prediction by comparing images of fluorescent neuronal morphology from nuclear-localized fluorescent protein to those from freely diffused fluorescent protein for classifying neuronal death. We found that biomarker-optimized (BO-) CNNs could learn to classify neuronal death from fluorescent protein-localized nuclear morphology (mApple-NLS-CNN) alone, with super-human accuracy. Furthermore, leveraging methods from explainable artificial intelligence, we identified novel features within the nuclear-localized fluorescent protein signal that were indicative of neuronal death. Our findings suggest that the use of a nuclear morphology marker in live imaging combined with computational models such mApple-NLS-CNN can provide an optimal readout of neuronal death, a common result of neurotoxicity.

Superhuman cell death detection with biomarker-optimized neural networks.

Cellular events underlying neurodegenerative disease may be captured by longitudinal live microscopy of neurons. While the advent of robot-assisted microscopy has helped scale such efforts to high-throughput regimes with the statistical power to detect transient events, time-intensive human annotation is required. We addressed this fundamental limitation with biomarker-optimized convolutional neural networks (BO-CNNs): interpretable computer vision models trained directly on biosensor activity. We demonstrate the ability of BO-CNNs to detect cell death, which is typically measured by trained annotators. BO-CNNs detected cell death with superhuman accuracy and speed by learning to identify subcellular morphology associated with cell vitality, despite receiving no explicit supervision to rely on these features. These models also revealed an intranuclear morphology signal that is difficult to spot by eye and had not previously been linked to cell death, but that reliably indicates death. BO-CNNs are broadly useful for analyzing live microscopy and essential for interpreting high-throughput experiments.

Optical force-induced nonlinearity and self-guiding of light in human red blood cell suspensions.

Osmotic conditions play an important role in the cell properties of human red blood cells (RBCs), which are crucial for the pathological analysis of some blood diseases such as malaria. Over the past decades, numerous efforts have mainly focused on the study of the RBC biomechanical properties that arise from the unique deformability of erythrocytes. Here, we demonstrate nonlinear optical effects from human RBCs suspended in different osmotic solutions. Specifically, we observe self-trapping and scattering-resistant nonlinear propagation of a laser beam through RBC suspensions under all three osmotic conditions, where the strength of the optical nonlinearity increases with osmotic pressure on the cells. This tunable nonlinearity is attributed to optical forces, particularly the forward-scattering and gradient forces. Interestingly, in aged blood samples (with lysed cells), a notably different nonlinear behavior is observed due to the presence of free hemoglobin. We use a theoretical model with an optical force-mediated nonlocal nonlinearity to explain the experimental observations. Our work on light self-guiding through scattering bio-soft-matter may introduce new photonic tools for noninvasive biomedical imaging and medical diagnosis.

Nonlinear Self-Action of Light through Biological Suspensions.

It is commonly thought that biological media cannot exhibit an appreciable nonlinear optical response. We demonstrate, for the first time to our knowledge, a tunable optical nonlinearity in suspensions of cyanobacteria that leads to robust propagation and strong self-action of a light beam. By deliberately altering the host environment of the marine bacteria, we show experimentally that nonlinear interaction can result in either deep penetration or enhanced scattering of light through the bacterial suspension, while the viability of the cells remains intact. A theoretical model is developed to show that a nonlocal nonlinearity mediated by optical forces (including both gradient and forward-scattering forces) acting on the bacteria explains our experimental observations.