Integrating deep learning and unbiased automated high-content screening to identify complex disease signatures in human fibroblasts

Abstract

Drug discovery for diseases such as Parkinson’s disease are impeded by the lack of screenable cellular phenotypes. We present an unbiased phenotypic profiling platform that combines automated cell culture, high-content imaging, Cell Painting, and deep learning. We applied this platform to primary fibroblasts from 91 Parkinson’s disease patients and matched healthy controls, creating the largest publicly available Cell Painting image dataset to date at 48 terabytes. We use fixed weights from a convolutional deep neural network trained on ImageNet to generate deep embeddings from each image and train machine learning models to detect morphological disease phenotypes. Our platform’s robustness and sensitivity allow the detection of individual-specific variation with high fidelity across batches and plate layouts. Lastly, our models confidently separate LRRK2 and sporadic Parkinson’s disease lines from healthy controls (receiver operating characteristic area under curve 0.79 (0.08 standard deviation)), supporting the capacity of this platform for complex disease modeling and drug screening applications.

Machine learning approaches based on fibroblast morphometry confidently identify stress but have limited ability to predict ALS

“…We formulated nine classification tasks and divided the fibroblasts into training and test sets (4:1 respectively on the cell line level). Dividing into sets on the line level is important to accurately measure model performance, because ML techniques were shown to recognize individual fibroblasts of the same line(44), and we saw inflated scores when set divisions were made on the single cell level. The tasks and the number of examples in the total, training, and test sets, broken down by class, are summarized in Table 1.…”

“…ML methods for different fluorescence microscopy tasks have been developing rapidly, including segmentation, object detection, denoising (46), image focus scoring (47), in Silico fluorescent labelling (48)(49)(50)(51), improvement of resolution( 52), label free cell death detection (53), and diagnosis of neurodegenerative diseases (40,44), in fibroblasts specifically. In this study, we used ML algorithms built on a large sample size and multiple features capable of capturing classes of environmental perturbations.…”

“…These limitations could reside in the choice of biological readouts. We focused on a number of parameters that we could feasibly obtain, were logically associated with ALS, and previously employed to develop biomarkers in fibroblasts for SMA (40) and Parkinson's disease (44). However, it could be necessary to test different sets of readouts, such as phosphorylated TDP-43 (21), ubiquitination (23), or nuclear pore proteins that are altered in ALS (54).…”

Induced Pluripotent Stem Cell-Based Drug Screening by Use of Artificial Intelligence

“…The training results showed that it was possible to separate fibroblasts derived from Parkinson’s disease from healthy controls. This is important data to show that the phenotype of the disease can be analyzed using AI [ 82 ].…”

A scalable screening platform for phenotypic subtyping of ALS patient-derived fibroblasts

“…However, mRNA alone may miss disease-relevant signals as ALS is also known to affect spatial organization at cell level, such as changes to protein localization and aggregation 22 . High-content microscopy can capture these spatial signals and has been applied in the context of Parkinson’s disease to classify disease vs. health 23 over a library of patient-derived fibroblasts. The use of fibroblasts is compelling, as fibroblasts are readily obtainable from living donors, retain the age of the patient, and can be scaled for high-throughput experimental assays.…”