A Systematic Evaluation of Interneuron Morphology Representations for Cell Type Discrimination
- PDF / 2,473,051 Bytes
- 19 Pages / 595.224 x 790.955 pts Page_size
- 54 Downloads / 163 Views
ORIGINAL ARTICLE
A Systematic Evaluation of Interneuron Morphology Representations for Cell Type Discrimination Sophie Laturnus1,4 · Dmitry Kobak1 · Philipp Berens1,2,3,4
© The Author(s) 2020
Abstract Quantitative analysis of neuronal morphologies usually begins with choosing a particular feature representation in order to make individual morphologies amenable to standard statistics tools and machine learning algorithms. Many different feature representations have been suggested in the literature, ranging from density maps to intersection profiles, but they have never been compared side by side. Here we performed a systematic comparison of various representations, measuring how well they were able to capture the difference between known morphological cell types. For our benchmarking effort, we used several curated data sets consisting of mouse retinal bipolar cells and cortical inhibitory neurons. We found that the best performing feature representations were two-dimensional density maps, two-dimensional persistence images and morphometric statistics, which continued to perform well even when neurons were only partially traced. Combining these feature representations together led to further performance increases suggesting that they captured non-redundant information. The same representations performed well in an unsupervised setting, implying that they can be suitable for dimensionality reduction or clustering. Keywords Neuroanatomy · Benchmarking · Cell types · Mouse · Visual cortex
Introduction The development of experimental methods for highthroughput single cell RNA sequencing (Zeisel et al. 2018; Saunders et al. 2018; Tasic et al. 2018; Cao et al. 2019) Electronic supplementary material The online version of this article (https://doi.org/10.1007/s12021-020-09461-z) contains supplementary material, which is available to authorized users. Philipp Berens
philipp.berens@uni-tuebingen.de Sophie Laturnus sophie.laturnus@uni-tuebingen.de Dmitry Kobak dmitry.kobak@uni-tuebingen.de 1
Institute for Ophthalmic Research, University of T¨ubingen, T¨ubingen, Germany
2
Institute for Bioinformatics and Medical Informatics, University of T¨ubingen, T¨ubingen, Germany
3
Bernstein Center for Computational Neuroscience, University of T¨ubingen, T¨ubingen, Germany
4
Center for Integrative Neuroscience, University of T¨ubingen, T¨ubingen, Germany
and large-scale functional imaging (Baden et al. 2016; Pachitariu et al. 2017; Schultz et al. 2017) has led to a surge of interest in identifying the building blocks of the brain – the neural cell types (Zeng and Sanes 2017; Xi et al. 2018). Both data modalities are analyzed with specialized quantitative tools (Stegle et al. 2015; Stringer and Pachitariu 2019) and produce data sets amenable to statistical analysis such as cell type identification by clustering. At the same time, ever since the work of Santiago Ram´on y Cajal (1899), it was the anatomy of a neuron that has been considered the defining feature of a neural cell type. Like in genetics and physiology, recent yea
