Automated Protein Crystallization Analysis Using Neural Networks

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Protein x-ray crystallography requires manual inspection of numerous protein crystallization trials, which is inefficient and time-consuming, as most trials do not result in crystallization, diverting researchers' efforts from analyzing protein structures.

Innovation Solution

An automated system using a camera and machine-learned neural networks to classify images of protein drops, identifying crystal formation and notifying users, with a taxonomy-based classification system and synthetic image augmentation for improved accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual inspection of protein crystallization trials is performed, then researchers can identify crystal formation, but researchers spend excessive time inspecting trials instead of analyzing protein structures

Engineering Contradiction:

Improvecrystal detection accuracyVSAvoidtime spent on trial inspection

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent replaces the manual mechanical inspection process with an automated image analysis system using machine learning models. The system captures images of crystallization trials and uses trained neural networks to automatically classify drops as containing crystals or not, eliminating the need for manual visual inspection while maintaining high detection accuracy.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system enables self-service by allowing the image analysis model to autonomously identify crystal formation without human intervention. The automated classification system processes images and generates results independently, freeing researchers from repetitive inspection tasks while preserving their ability to review and validate results when needed.

Inventive Principle:

Principle #25Self-service

2Adaptability or versatility

If large numbers of crystallization trials are conducted under wide range of conditions, then comprehensive protein crystallization data is obtained, but the complexity of managing and analyzing these trials increases

Engineering Contradiction:

Improvecrystallization condition coverageVSAvoidtrial management complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The image analysis system serves multiple functions: it classifies crystal formation, quantifies crystal characteristics, and provides structured output data. This multi-functional approach handles diverse trial conditions uniformly, managing the complexity of wide-ranging crystallization experiments through a single versatile automated system rather than multiple specialized manual processes.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS11010591B2Automatic protein crystallization trial analysis system

Publication Date: 2021.05.18 MERCK SHARP & DOHME LLC

AI summary

A protein crystallization trial is automatically analyzed by capturing images of the protein drops in the trial. A machine-learned model, such as a neural network, is applied to classify the images. The model generates a predicted classification from among a set of possible classifications which includes one or more crystal type classifications and one or more non-crystal type classifications. Users may be notified automatically of newly identified crystals (e.g., drops that are classified as a crystal type). The notification may include a link to a user interface that includes results of the trial.