HCS-PhenoCluster - Machine Learning based High-Content Phenotypic Analysis Tool

Computational analysis of multi-parametric high-content imaging dataset can be particularly tedious and daunting to implement for a laboratory biologist with no bioinformatics expertise and hence, there is an unmet need of software tools to facilitate this type of analysis. We present a web-based application HCS-PhenoCluster for the analysis of high-content image-based data which can be used for discovering novel cellular phenotypes beyond visual inspection. This application is interfaced with a single-cell extraction pipeline which performs cell segmentation and multi-feature extraction of high-content imaging data. HCS-PhenoCluster has been implemented in Swift language using Vapor framework and consists of a MySQL database. The image analysis workflow of HCS-PhenoCluster is based on machine learning models implemented in Python and comprises five modules of data processing which include multi-level quality control modules and an unsupervised clustering module to reveal phenotypic diversity in the dataset. We present a case study of the application of this tool for detecting multiple Golgi organizational states. The user manual including the description of methodology and relevant references are available with the application.