A spectral graph theoretic approach to quantification and calibration of collective morphological differences in cell images

Abstract

Motivation: High-throughput image-based assay technologies can rapidly produce a large number of cell images for drug screening, but data analysis is still a major bottleneck that limits their utility. Quantifying a wide variety of morphological differences observed in cell images under different drug influences is still a challenging task because the result can be highly sensitive to sampling and noise.Results: We propose a graph-based approach to cell image analysis. We define graph transition energy to quantify morphological differences between image sets. A spectral graph theoretic regularization is applied to transform the feature space based on training examples of extremely different images to calibrate the quantification. Calibration is essential for a practical quantification method because we need to measure the confidence of the quantification. We applied our method to quantify the degree of partial fragmentation of mitochondria in collections of fluorescent cell images. We show that with transformation, the quantification can be more accurate and sensitive than that without transformation. We also show that our method outperforms competing methods, including neighbourhood component analysis and the multi-variate drug profiling method by Loo et al. We illustrate its utility with a study of Annonaceous acetogenins, a family of compounds with drug potential. Our result reveals that squamocin induces more fragmented mitochondria than muricin A.Availability: Mitochondrial cell images, their corresponding feature sets (SSLF and WSLF) and the source code of our proposed method are available at http://aiia.iis.sinica.edu.tw/.Contact: chunnan@iis.sinica.edu.twSupplementary information: Supplementary data are available at Bioinformatics online.

MitoLoc: A method for the simultaneous quantification of mitochondrial network morphology and membrane potential in single cells

“…The characterisation of mitochondrial volume and morphology is thus easily biassed and dependent on the experience of the experimenter. Recently however, elaborate computational analysis of mitochondrial structures has shown that mitochondrial networks can in principle be described in numerical terms ( Lin et al, 2010; Peng et al, 2011; Sukhorukov et al, 2012 ). If easily available, such metrics would thus reduce the subjectivity in assessing mitochondrial morphologies.…”

Automatic quantification of mitochondrial fragmentation from two‐photon microscope images of mouse brain tissue

“…Next, we compared the method to that proposed in (Lin et al., ) for quantifying mitochondrial fragmentation from images of single cells. Their method uses a different set of features, and the fragmentation is quantified using weighted 6‐nearest neighbour regression.…”

“…The predicted score for the nonfragmented mitochondria is overestimated, and the predicted score for the fragmented mitochondria is underestimated; however, this bias is least apparent in the case of MT + TF. Next, we compared the method to that proposed in (Lin et al, 2010) for quantifying mitochondrial fragmentation from images of single cells. Their method uses a different set of features, and the fragmentation is quantified using weighted 6-nearest neighbour regression.…”

“…Here, we first describe our approach to quantifying mitochondrial fragmentation, and evaluate its performance. Next, we compare the performance of our method to previous ones [namely, the one proposed in (Lin et al, 2010), which also does not require use of reference images]. In addition, we test our method on synthetic data.…”

Informative Energy Metric for Similarity Measure in Reproducing Kernel Hilbert Spaces

“…The main idea is that we quantify the amount of information between particles according to their graph energy 18 . Our goal is to transform the kernel space so that the distance in the transformed space correlated with the difference of the labels of particles.…”