In Silico Labeling: Predicting Fluorescent Labels in Unlabeled Images

Abstract

Microscopy is a central method in life sciences. Many popular methods, such as antibody labeling, are used to add physical fluorescent labels to specific cellular constituents. However, these approaches have significant drawbacks, including inconsistency; limitations in the number of simultaneous labels because of spectral overlap; and necessary perturbations of the experiment, such as fixing the cells, to generate the measurement. Here, we show that a computational machine-learning approach, which we call "in silico labeling" (ISL), reliably predicts some fluorescent labels from transmitted-light images of unlabeled fixed or live biological samples. ISL predicts a range of labels, such as those for nuclei, cell type (e.g., neural), and cell state (e.g., cell death). Because prediction happens in silico, the method is consistent, is not limited by spectral overlap, and does not disturb the experiment. ISL generates biological measurements that would otherwise be problematic or impossible to acquire.

Deep learning of diffraction image patterns for accurate classification of five cell types

“…Rapid classification of cells according to their phenotypes and/or pathologic conditions is a significant and long‐standing problem . While genomic analysis offers the gold standard , accurate and cost‐effective methods at the single‐cell level are highly valuable and attract intense research efforts . Conventional methods of microscopy and flow cytometry are current tools of choice, but the former requires time‐consuming manual analysis and the latter yields very limited morphology information.…”

Internet of instruments

“…Use of image processing algorithms can only partly overcome this deficiency, so there has been interest in developing methods that avoid the use of fluorescence altogether. Several US institutions therefore developed a ML approach called in silico labeling that can predict reliably where fluorescent labels would be applied based on images of the either unlabeled fixed or live biological samples [2]. The researchers designed a model based on deep neural networks, which comprise nodes and links between them and are arranged in multiple layers to allow nested feedback between data and predictions.…”

Deep learning to predict microscope images

“…Two recent studies show how the breathtakingly rapid evolution of methods in Machine Learning (ML) might change that situation. Work by Christiansen et al 1 , and Ounkomol et al 2 ., exploits not-yet-mainstream developments in ML to train neural networks (NNs) to harvest this information and make it available to humans.…”

“…Christiansen et al 1 . and Ounkomol et al 2 extend this splendid artificiality to carry out a task quite different from classification.…”

“…Use of NNs to predict fluorescence may turn out to be a case in which recondite computational methods, first developed for military purposes by a nation state, extended by large corporations 1 and institutions 2 in coastal enclaves, ultimately results in worldwide democratization of technical capability and a burst of new knowledge. That outcome isn’t preordained.…”