In-Vitro Microscopy Sample QC Using Z-Slice Viability Screening
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional semi-automated toxicity prediction assays in high throughput screening (HTS) are unreliable in identifying the non-toxicity or toxicity of compounds, leading to a high risk of performing in-vivo trials with compounds that have passed these assays, particularly in the case of drug-induced liver injury.
Innovation Solution
A computer-implemented method using machine learning models, including a convolutional neural network (CNN) and one-class SVM, to identify viable samples from HTS data, followed by a UMAP or t-SNE algorithm for dimensional reduction and a Wasserstein distance metric to predict toxicity or non-toxicity, enhancing the reliability of downstream analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional semi-automated test systems and methodologies are used to identify compounds that produce a detectable signal, then the testing process can be conducted with standard equipment, but the reliability of identifying non-toxicity or toxicity of compounds is insufficient
Solution Approach 1:
The quality control process is segmented into multiple independent analysis steps: generating 2D image slices along the z-axis, identifying viable samples from image slices, detecting artifacts and erroneous data points, and performing dimensional reduction. Each segment can be independently optimized and validated, improving overall reliability while maintaining manageable system complexity.
Solution Approach 2:
The patent transforms 3D cellular structure data into multiple 2D image slices along the z-axis, enabling analysis from different dimensional perspectives. This dimensional transformation allows for more comprehensive artifact detection and viability assessment without requiring complex 3D analysis algorithms, thus improving reliability while controlling complexity.
2Productivity
If High Throughput Screening (HTS) is used to test a large number of potential compounds, then the screening speed and throughput increase, but the large amount of data generated requires careful analysis to detect artifacts and correct erroneous data points
Solution Approach 1:
The system performs preliminary quality control actions automatically during the HTS process: generating 2D image slices, identifying viable samples, and detecting artifacts before downstream analysis. This preliminary filtering reduces the burden on subsequent data analysis steps and maintains high throughput by preventing error propagation.
Solution Approach 2:
The quality control system is self-service in that it automatically identifies viable samples, detects artifacts, and corrects erroneous data points without requiring manual intervention. The machine learning models and algorithms autonomously process the large HTS datasets, maintaining productivity while managing analysis complexity through automation.
3Reliability
If manual analysis by researchers is performed to detect artifacts and correct erroneous data points, then data quality can be improved, but the time required for analysis increases and productivity decreases
Solution Approach 1:
The patent replaces manual mechanical analysis by researchers with automated computational systems. Machine learning models and algorithms automatically perform artifact detection and error correction, achieving comparable or superior data quality while dramatically reducing analysis time. This substitution maintains reliability through consistent automated application of quality control criteria.
4Productivity
If compounds that have passed semi-automated assays are advanced to in-vivo trials, then the drug development process can proceed efficiently, but the risk of performing trials with compounds that are actually toxic increases
Solution Approach 1:
The enhanced quality control system performs preliminary toxicity assessment by automatically identifying viable samples and detecting artifacts before compounds advance to in-vivo trials. This preliminary filtering action reduces the risk of advancing toxic compounds while maintaining drug development efficiency through automated high-throughput analysis.
Solution Approach 2:
The system incorporates feedback mechanisms where quality control metrics from 2D image slice analysis inform downstream toxicity prediction models. This feedback loop continuously refines the identification of viable versus non-viable samples, improving the reliability of toxicity predictions and reducing false positives that could lead to unsafe in-vivo trials.
Data Source
AI summary
Methods, apparatus, systems and computer-implemented methods configured for identifying viable samples of cellular structures for analysis in an in-vitro microscopy assay. Automatically identifying a first set of samples useful for analysis from a plurality of samples of an assay plate. Generating a set of 2-dimensional (2D) images for each sample in the first set of samples. The set of 2D images for said each sample comprising multiple 2D image slices taken along a z-axis of said each sample. Identifying from the sets of 2D image slices a set of viable samples. Outputting data representative of said set of viable samples for analysis as the set of images.