Deep learning analysis of mid-infrared microscopic imaging data for the diagnosis and classification of human lymphomas

The present study presents an alternative analytical workflow that combines mid-infrared (MIR) microscopic imaging and deep learning to diagnose human lymphoma and differentiate between small and large cell lymphoma. We could show that using a deep learning approach to analyze MIR hyperspectral data obtained from benign and malignant lymph node pathology results in high accuracy for correct classification, learning the distinct region of 3900 to 850 cm-1 . The accuracy is above 95% for every pair of malignant lymphoid tissue and still above 90% for the distinction between benign and malignant lymphoid tissue for binary classification. These results demonstrate that a preliminary diagnosis and subtyping of human lymphoma could be streamlined by applying a deep learning approach to analyze MIR spectroscopic data.

The Impact of Tissue Preparation on Salivary Gland Tumors Investigated by Fourier-Transform Infrared Microspectroscopy

Due to the wide variety of benign and malignant salivary gland tumors, classification and malignant behavior determination based on histomorphological criteria can be difficult and sometimes impossible. Spectroscopical procedures can acquire molecular biological information without destroying the tissue within the measurement processes. Since several tissue preparation procedures exist, our study investigated the impact of these preparations on the chemical composition of healthy and tumorous salivary gland tissue by Fourier-transform infrared (FTIR) microspectroscopy. Sequential tissue cross-sections were prepared from native, formalin-fixed and formalin-fixed paraffin-embedded (FFPE) tissue and analyzed. The FFPE cross-sections were dewaxed and remeasured. By using principal component analysis (PCA) combined with a discriminant analysis (DA), robust models for the distinction of sample preparations were built individually for each parotid tissue type. As a result, the PCA-DA model evaluation showed a high similarity between native and formalin-fixed tissues based on their chemical composition. Thus, formalin-fixed tissues are highly representative of the native samples and facilitate a transfer from scientific laboratory analysis into the clinical routine due to their robust nature. Furthermore, the dewaxing of the cross-sections entails the loss of molecular information. Our study successfully demonstrated how FTIR microspectroscopy can be used as a powerful tool within existing clinical workflows.

Mito Hacker: a set of tools to enable high-throughput analysis of mitochondrial network morphology

Mitochondria are highly dynamic organelles that can exhibit a wide range of morphologies. Mitochondrial morphology can differ significantly across cell types, reflecting different physiological needs, but can also change rapidly in response to stress or the activation of signaling pathways. Understanding both the cause and consequences of these morphological changes is critical to fully understanding how mitochondrial function contributes to both normal and pathological physiology. However, while robust and quantitative analysis of mitochondrial morphology has become increasingly accessible, there is a need for new tools to generate and analyze large data sets of mitochondrial images in high throughput. The generation of such datasets is critical to fully benefit from rapidly evolving methods in data science, such as neural networks, that have shown tremendous value in extracting novel biological insights and generating new hypotheses. Here we describe a set of three computational tools, Cell Catcher, Mito Catcher and MiA, that we have developed to extract extensive mitochondrial network data on a single-cell level from multi-cell fluorescence images. Cell Catcher automatically separates and isolates individual cells from multi-cell images; Mito Catcher uses the statistical distribution of pixel intensities across the mitochondrial network to detect and remove background noise from the cell and segment the mitochondrial network; MiA uses the binarized mitochondrial network to perform more than 100 mitochondria-level and cell-level morphometric measurements. To validate the utility of this set of tools, we generated a database of morphological features for 630 individual cells that encode 0, 1 or 2 alleles of the mitochondrial fission GTPase Drp1 and demonstrate that these mitochondrial data could be used to predict Drp1 genotype with 87% accuracy. Together, this suite of tools enables the high-throughput and automated collection of detailed and quantitative mitochondrial structural information at a single-cell level. Furthermore, the data generated with these tools, when combined with advanced data science approaches, can be used to generate novel biological insights.

The Cell Painting Assay as a Screening Tool for the Discovery of Bioactivities in New Chemical Matter

Multiparametric phenotypic screening based on cellular morphology interrogates many biological pathways simultaneously and is therefore a valuable screening tool for the discovery of new biological activities. The cell painting assay stains various cellular features using six different dyes in one well. By automated image analysis, hundreds of parameters are calculated from the images which deliver a phenotypic profile of the cell. It has been shown that compounds with similar modes of action deliver similar phenotypic profiles. Using a reference set of compounds with known modes of action, it is possible to assign probable modes of action to new compounds and to discover compounds with potentially new modes of action.Here we describe the cell painting assay as a screening tool using a hit identification workflow which has been implemented using open-source software.

RefCell: multi-dimensional analysis of image-based high-throughput screens based on 'typical cells'

BACKGROUND

Image-based high-throughput screening (HTS) reveals a high level of heterogeneity in single cells and multiple cellular states may be observed within a single population. Currently available high-dimensional analysis methods are successful in characterizing cellular heterogeneity, but suffer from the "curse of dimensionality" and non-standardized outputs.

RESULTS

Here we introduce RefCell, a multi-dimensional analysis pipeline for image-based HTS that reproducibly captures cells with typical combinations of features in reference states and uses these "typical cells" as a reference for classification and weighting of metrics. RefCell quantitatively assesses heterogeneous deviations from typical behavior for each analyzed perturbation or sample.

CONCLUSIONS

We apply RefCell to the analysis of data from a high-throughput imaging screen of a library of 320 ubiquitin-targeted siRNAs selected to gain insights into the mechanisms of premature aging (progeria). RefCell yields results comparable to a more complex clustering-based single-cell analysis method; both methods reveal more potential hits than a conventional analysis based on averages.

Utilization of Multidimensional Data in the Analysis of Ultra-High-Throughput High Content Phenotypic Screens

High Content Screening (HCS) platforms can generate large amounts of multidimensional data. To take full advantage of all the rich contextual information provided by these screens, a combination of traditional as well as nontraditional hit identification and prioritization strategies is required. Here, we describe the workflow and analytics of multidimensional high content data to differentiate, group, and prioritize hits.

Effect-Size Measures as Descriptors of Assay Quality in High-Content Screening: A Brief Review of Some Available Methodologies

The field of high-content screening (HCS) typically uses measures of screen quality conceived for fairly straightforward high-throughput screening (HTS) scenarios. However, in contrast to HTS, image-based HCS systems rely on multidimensional readouts reporting biological responses associated with complex cellular phenotypes. Not only is the dimensionality in which the screens operate higher, but also the scale of the individual features describing the quantified phenotypic changes is often smaller than what is seen in one-dimensional HTS platforms. Therefore, the use of HTS-type quality measures to characterize HCS screens may severely underestimate the detection power of the assays used, and it may mislead the screening scientists regarding the necessary screen optimization. Also, traditional HTS-based measures are typically reported without any estimation of precision, which makes them unsuitable for computation of confidence intervals or for meta-analysis. This review summarizes the well-established statistical techniques for reporting effect sizes and argues that this broadly accepted methodology could be seamlessly integrated into HCS quality reporting, supplanting or even replacing measures such as Z' or S_w (assay window).

Applications in image-based profiling of perturbations

A dramatic shift has occurred in how biologists use microscopy images. Whether experiments are small-scale or high-throughput, automatically quantifying biological properties in images is now widespread. We see yet another revolution under way: a transition towards using automated image analysis to not only identify phenotypes a biologist specifically seeks to measure ('screening') but also as an unbiased and sensitive tool to capture a wide variety of subtle features of cell (or organism) state ('profiling'). Mapping similarities among samples using image-based (morphological) profiling has tremendous potential to transform drug discovery, functional genomics, and basic biological research. Applications include target identification, lead hopping, library enrichment, functionally annotating genes/alleles, and identifying small molecule modulators of gene activity and disease-specific phenotypes.

Toward high-content screening of mitochondrial morphology and membrane potential in living cells

Mitochondria are double membrane organelles involved in various key cellular processes. Governed by dedicated protein machinery, mitochondria move and continuously fuse and divide. These "mitochondrial dynamics" are bi-directionally linked to mitochondrial and cell functional state in space and time. Due to the action of the electron transport chain (ETC), the mitochondrial inner membrane displays a inside-negative membrane potential (Δψ). The latter is considered a functional readout of mitochondrial "health" and required to sustain normal mitochondrial ATP production and mitochondrial fusion. During the last decade, live-cell microscopy strategies were developed for simultaneous quantification of Δψ and mitochondrial morphology. This revealed that ETC dysfunction, changes in Δψ and aberrations in mitochondrial structure often occur in parallel, suggesting they are linked potential targets for therapeutic intervention. Here we discuss how combining high-content and high-throughput strategies can be used for analysis of genetic and/or drug-induced effects at the level of individual organelles, cells and cell populations. This article is part of a Directed Issue entitled: Energy Metabolism Disorders and Therapies.

Image-based cell-resolved screening assays in flow

A parallel microfluidic cytometer (PMC) is based on a one-dimensional (1D) scanning detector, a parallel array of flow channels, and new multiparameter analysis algorithms that operate on low-pixel-count 1D images. In this article, we explore a series of image-based live- and fixed-cell screening assays, including two NF-kB nuclear translocations and T-cell capping. We then develop a new multiparametric linear weighted classifier that achieves a Z' factor sufficient for scaled pharmaceutical discovery with Jurkat cells in suspension. We conclude that the PMC should have the throughput and statistical power to permit a new capability for image-based high-sample-number pharmaceutical screening with suspension samples.

Workflow for high-content, individual cell quantification of fluorescent markers from universal microscope data, supported by open source software

Advances in understanding the control mechanisms governing the behavior of cells in adherent mammalian tissue culture models are becoming increasingly dependent on modes of single-cell analysis. Methods which deliver composite data reflecting the mean values of biomarkers from cell populations risk losing subpopulation dynamics that reflect the heterogeneity of the studied biological system. In keeping with this, traditional approaches are being replaced by, or supported with, more sophisticated forms of cellular assay developed to allow assessment by high-content microscopy. These assays potentially generate large numbers of images of fluorescent biomarkers, which enabled by accompanying proprietary software packages, allows for multi-parametric measurements per cell. However, the relatively high capital costs and overspecialization of many of these devices have prevented their accessibility to many investigators. Described here is a universally applicable workflow for the quantification of multiple fluorescent marker intensities from specific subcellular regions of individual cells suitable for use with images from most fluorescent microscopes. Key to this workflow is the implementation of the freely available Cell Profiler software(1) to distinguish individual cells in these images, segment them into defined subcellular regions and deliver fluorescence marker intensity values specific to these regions. The extraction of individual cell intensity values from image data is the central purpose of this workflow and will be illustrated with the analysis of control data from a siRNA screen for G1 checkpoint regulators in adherent human cells. However, the workflow presented here can be applied to analysis of data from other means of cell perturbation (e.g., compound screens) and other forms of fluorescence based cellular markers and thus should be useful for a wide range of laboratories.

A comparative study of cell classifiers for image-based high-throughput screening

Background

Millions of cells are present in thousands of images created in high-throughput screening (HTS). Biologists could classify each of these cells into a phenotype by visual inspection. But in the presence of millions of cells this visual classification task becomes infeasible. Biologists train classification models on a few thousand visually classified example cells and iteratively improve the training data by visual inspection of the important misclassified phenotypes. Classification methods differ in performance and performance evaluation time. We present a comparative study of computational performance of gentle boosting, joint boosting CellProfiler Analyst (CPA), support vector machines (linear and radial basis function) and linear discriminant analysis (LDA) on two data sets of HT29 and HeLa cancer cells.

Results

For the HT29 data set we find that gentle boosting, SVM (linear) and SVM (RBF) are close in performance but SVM (linear) is faster than gentle boosting and SVM (RBF). For the HT29 data set the average performance difference between SVM (RBF) and SVM (linear) is 0.42 %. For the HeLa data set we find that SVM (RBF) outperforms other classification methods and is on average 1.41 % better in performance than SVM (linear).

Conclusions

Our study proposes SVM (linear) for iterative improvement of the training data and SVM (RBF) for the final classifier to classify all unlabeled cells in the whole data set.

Electronic supplementary material

The online version of this article (doi:10.1186/1471-2105-15-342) contains supplementary material, which is available to authorized users.

Cell-based fuzzy metrics enhance high-content screening (HCS) assay robustness

High-content screening (HCS) allows the exploration of complex cellular phenotypes by automated microscopy and is increasingly being adopted for small interfering RNA genomic screening and phenotypic drug discovery. We introduce a series of cell-based evaluation metrics that have been implemented and validated in a mono-parametric HCS for regulators of the membrane trafficking protein caveolin 1 (CAV1) and have also proved useful for the development of a multiparametric phenotypic HCS for regulators of cytoskeletal reorganization. Imaging metrics evaluate imaging quality such as staining and focus, whereas cell biology metrics are fuzzy logic-based evaluators describing complex biological parameters such as sparseness, confluency, and spreading. The evaluation metrics were implemented in a data-mining pipeline, which first filters out cells that do not pass a quality criterion based on imaging metrics and then uses cell biology metrics to stratify cell samples to allow further analysis of homogeneous cell populations. Use of these metrics significantly improved the robustness of the monoparametric assay tested, as revealed by an increase in Z' factor, Kolmogorov-Smirnov distance, and strict standard mean difference. Cell biology evaluation metrics were also implemented in a novel supervised learning classification method that combines them with phenotypic features in a statistical model that exceeded conventional classification methods, thus improving multiparametric phenotypic assay sensitivity.

Benchmarking of Multivariate Similarity Measures for High-Content Screening Fingerprints in Phenotypic Drug Discovery

High-content screening (HCS) is a powerful tool for drug discovery being capable of measuring cellular responses to chemical disturbance in a high-throughput manner. HCS provides an image-based readout of cellular phenotypes, including features such as shape, intensity, or texture in a highly multiplexed and quantitative manner. The corresponding feature vectors can be used to characterize phenotypes and are thus defined as HCS fingerprints. Systematic analyses of HCS fingerprints allow for objective computational comparisons of cellular responses. Such comparisons therefore facilitate the detection of different compounds with different phenotypic outcomes from high-throughput HCS campaigns. Feature selection methods and similarity measures, as a basis for phenotype identification and clustering, are critical for the quality of such computational analyses. We systematically evaluated 16 different similarity measures in combination with linear and nonlinear feature selection methods for their potential to capture biologically relevant image features. Nonlinear correlation-based similarity measures such as Kendall’s τ and Spearman’s ρ perform well in most evaluation scenarios, outperforming other frequently used metrics (such as the Euclidian distance). We also present four novel modifications of the connectivity map similarity that surpass the original version, in our experiments. This study provides a basis for generic phenotypic analysis in future HCS campaigns.

Comparison of methods for image-based profiling of cellular morphological responses to small-molecule treatment

Quantitative microscopy has proven a versatile and powerful phenotypic screening technique. Recently, image-based profiling has shown promise as a means for broadly characterizing molecules' effects on cells in several drug-discovery applications, including target-agnostic screening and predicting a compound's mechanism of action (MOA). Several profiling methods have been proposed, but little is known about their comparative performance, impeding the wider adoption and further development of image-based profiling. We compared these methods by applying them to a widely applicable assay of cultured cells and measuring the ability of each method to predict the MOA of a compendium of drugs. A very simple method that is based on population means performed as well as methods designed to take advantage of the measurements of individual cells. This is surprising because many treatments induced a heterogeneous phenotypic response across the cell population in each sample. Another simple method, which performs factor analysis on the cellular measurements before averaging them, provided substantial improvement and was able to predict MOA correctly for 94% of the treatments in our ground-truth set. To facilitate the ready application and future development of image-based phenotypic profiling methods, we provide our complete ground-truth and test data sets, as well as open-source implementations of the various methods in a common software framework.

Finding novel pharmaceuticals in the systems biology era using multiple effective drug targets, phenotypic screening and knowledge of transporters: where drug discovery went wrong and how to fix it

Despite the sequencing of the human genome, the rate of innovative and successful drug discovery in the pharmaceutical industry has continued to decrease. Leaving aside regulatory matters, the fundamental and interlinked intellectual issues proposed to be largely responsible for this are: (a) the move from 'function-first' to 'target-first' methods of screening and drug discovery; (b) the belief that successful drugs should and do interact solely with single, individual targets, despite natural evolution's selection for biochemical networks that are robust to individual parameter changes; (c) an over-reliance on the rule-of-5 to constrain biophysical and chemical properties of drug libraries; (d) the general abandoning of natural products that do not obey the rule-of-5; (e) an incorrect belief that drugs diffuse passively into (and presumably out of) cells across the bilayers portions of membranes, according to their lipophilicity; (f) a widespread failure to recognize the overwhelmingly important role of proteinaceous transporters, as well as their expression profiles, in determining drug distribution in and between different tissues and individual patients; and (g) the general failure to use engineering principles to model biology in parallel with performing 'wet' experiments, such that 'what if?' experiments can be performed in silico to assess the likely success of any strategy. These facts/ideas are illustrated with a reasonably extensive literature review. Success in turning round drug discovery consequently requires: (a) decent systems biology models of human biochemical networks; (b) the use of these (iteratively with experiments) to model how drugs need to interact with multiple targets to have substantive effects on the phenotype; (c) the adoption of polypharmacology and/or cocktails of drugs as a desirable goal in itself; (d) the incorporation of drug transporters into systems biology models, en route to full and multiscale systems biology models that incorporate drug absorption, distribution, metabolism and excretion; (e) a return to 'function-first' or phenotypic screening; and (f) novel methods for inferring modes of action by measuring the properties on system variables at all levels of the 'omes. Such a strategy offers the opportunity of achieving a state where we can hope to predict biological processes and the effect of pharmaceutical agents upon them. Consequently, this should both lower attrition rates and raise the rates of discovery of effective drugs substantially.

Fluorescence-based high-throughput functional profiling of ligand-gated ion channels at the level of single cells

Ion channels are involved in many physiological processes and are attractive targets for therapeutic intervention. Their functional properties vary according to their subunit composition, which in turn varies in a developmental and tissue-specific manner and as a consequence of pathophysiological events. Understanding this diversity requires functional analysis of ion channel properties in large numbers of individual cells. Functional characterisation of ligand-gated channels involves quantitating agonist and drug dose-response relationships using electrophysiological or fluorescence-based techniques. Electrophysiology is limited by low throughput and high-throughput fluorescence-based functional evaluation generally does not enable the characterization of the functional properties of each individual cell. Here we describe a fluorescence-based assay that characterizes functional channel properties at single cell resolution in high throughput mode. It is based on progressive receptor activation and iterative fluorescence imaging and delivers >100 dose-responses in a single well of a 384-well plate, using α1-3 homomeric and αβ heteromeric glycine receptor (GlyR) chloride channels as a model system. We applied this assay with transiently transfected HEK293 cells co-expressing halide-sensitive yellow fluorescent protein and different GlyR subunit combinations. Glycine EC₅₀ values of different GlyR isoforms were highly correlated with published electrophysiological data and confirm previously reported pharmacological profiles for the GlyR inhibitors, picrotoxin, strychnine and lindane. We show that inter and intra well variability is low and that clustering of functional phenotypes permits identification of drugs with subunit-specific pharmacological profiles. As this method dramatically improves the efficiency with which ion channel populations can be characterized in the context of cellular heterogeneity, it should facilitate systems-level analysis of ion channel properties in health and disease and the discovery of therapeutics to reverse pathological alterations.

Shedding light on filovirus infection with high-content imaging

Microscopy has been instrumental in the discovery and characterization of microorganisms. Major advances in high-throughput fluorescence microscopy and automated, high-content image analysis tools are paving the way to the systematic and quantitative study of the molecular properties of cellular systems, both at the population and at the single-cell level. High-Content Imaging (HCI) has been used to characterize host-virus interactions in genome-wide reverse genetic screens and to identify novel cellular factors implicated in the binding, entry, replication and egress of several pathogenic viruses. Here we present an overview of the most significant applications of HCI in the context of the cell biology of filovirus infection. HCI assays have been recently implemented to quantitatively study filoviruses in cell culture, employing either infectious viruses in a BSL-4 environment or surrogate genetic systems in a BSL-2 environment. These assays are becoming instrumental for small molecule and siRNA screens aimed at the discovery of both cellular therapeutic targets and of compounds with anti-viral properties. We discuss the current practical constraints limiting the implementation of high-throughput biology in a BSL-4 environment, and propose possible solutions to safely perform high-content, high-throughput filovirus infection assays. Finally, we discuss possible novel applications of HCI in the context of filovirus research with particular emphasis on the identification of possible cellular biomarkers of virus infection.