Automated Learning of Subcellular Variation among Punctate Protein Patterns and a Generative Model of Their Relation to Microtubules.

Gregory R Johnson,Robert F Murphy,Gustavo K Rohde,Aabid Shariff,Jieyue Li,Achim Tresch

doi:10.1371/journal.pcbi.1004614

Gregory R Johnson, Robert F Murphy + Show 4 more

Open Access

https://doi.org/10.1371/journal.pcbi.1004614

Copy DOI

Abstract

Characterizing the spatial distribution of proteins directly from microscopy images is a difficult problem with numerous applications in cell biology (e.g. identifying motor-related proteins) and clinical research (e.g. identification of cancer biomarkers). Here we describe the design of a system that provides automated analysis of punctate protein patterns in microscope images, including quantification of their relationships to microtubules. We constructed the system using confocal immunofluorescence microscopy images from the Human Protein Atlas project for 11 punctate proteins in three cultured cell lines. These proteins have previously been characterized as being primarily located in punctate structures, but their images had all been annotated by visual examination as being simply “vesicular”. We were able to show that these patterns could be distinguished from each other with high accuracy, and we were able to assign to one of these subclasses hundreds of proteins whose subcellular localization had not previously been well defined. In addition to providing these novel annotations, we built a generative approach to modeling of punctate distributions that captures the essential characteristics of the distinct patterns. Such models are expected to be valuable for representing and summarizing each pattern and for constructing systems biology simulations of cell behaviors.

Highlights

Fluorescence microscope images can provide important information about the subcellular location of proteins, and automated systems can be used to assign these proteins to major subcellular location classes with accuracy at or above that of human annotators [1, 2]
Fluorescence microscopy is the main source of information about subcellular location, but large collections of fluorescence images for many proteins are frequently annotated visually and result in assignment only to broad categories
We describe automated methods for analyzing images from the Human Protein Atlas to identify nine specific punctate patterns and assign these more specific annotations to 550 proteins many of which previously had little information about subcellular location

Summary

Introduction

Fluorescence microscope images can provide important information about the subcellular location of proteins, and automated systems can be used to assign these proteins to major subcellular location classes with accuracy at or above that of human annotators [1, 2]. Assigning higher resolution annotations to proteins is more difficult, especially for punctate or vesicular patterns. Punctate subcellular localization patterns may arise either from membranebound organelles (e.g., transport vesicles) or from macromolecular complexes of sufficient size (e.g., ribonucleoprotein (RNP) bodies), and they may be quite visually similar. We refer to individual components of these patterns collectively as puncta, to encompass both types of structures. These are important for various cellular tasks such as endocytosis, exocytosis and RNA recruitment, storage or degradation. A critical factor for accomplishing many of those tasks is the association of the vesicles or bodies with cytoskeletal components such as microtubules for intracellular transport. The extent to which the distributions of specific puncta are related to that of microtubules remains unclear, as is the extent to which the distributions vary across different cell lines

Methods

Results

Discussion

Conclusion