High-throughput behavioral screening in Caenorhabditis elegans using machine learning for drug repurposing

Aug 3, 2026

Upload by Ophelia

Caenorhabditis elegans serves as an animal model for studying new disease treatments, yet conventional automated phenotyping can miss subtle, non-linear mobility patterns. This work proposes a high-throughput machine-learning screening framework that uses classifier outputs representing a recovery percentage as a quantitative treatment-effect measure. Two pipelines are evaluated: traditional ML with Tierpsy Tracker skeleton-derived behavioral features, and deep neural networks operating directly on video sequences. Results show that a Random Forest model yields higher accuracy and better explainability than deep learning, and application to a published unc-80 mutant drug-repurposing dataset demonstrates improved robustness for complex phenotypic differences.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages13

Reading Time33 min

AccessFree to download

Introduction

Motivation from drug discovery and disease models

Video-based phenotyping and automated feature extraction

Prior drug repurposing approach and its limitations

Statistical feature testing with multiple-comparison correction

Why subtle patterns may be missed

Proposed machine-learning screening framework

Recovery-percentage classifier output as treatment-effect metric

Feature-based ML using Tierpsy Tracker

Deep neural networks from video sequences

Evaluation and application

Comparative results and model suitability

Case study on unc-80 mutant repurposing data

[www. nature.com/scientificreports](www. nature.com/scientificreports) OPEN High-throughput behavioral screening in Caenorhabditis elegans using machine learning for drug repurposing Antonio García-Garví & Antonio-José Sánchez-Salmerón􀀍 Caenorhabditis elegans is a widely used animal model for researching new disease treatments. In recent years, automated methods have been developed to extract mobility phenotypes and analyse, using statistical methods, whether there are differences between control strains and disease modelstrains. However, these methods present certain limitations in detecting subtle and non-linear patterns. In this study, we propose a high-throughput screening method based on machine learning, using classifiers that provide a recovery percentage as a measure of treatment effect. We evaluate two main approaches: traditional machine learning models based on behavioral features extracted from the worm’s skeleton using Tierpsy Tracker, and deep neural networks that directly analyse video sequences. The results indicate that a Random Forest classifier trained with features extracted by Tierpsy Tracker offers higher accuracy and explainability, making it more suitable than deep learning models for drug testing experiments. Finally, to assess the applicability of our method, we processed data from a published drug repurposing study on unc-80 mutants based on statistical methods. The results highlight the potential of machine learning models to enhance automated phenotypic screening in animal models, providing a more robust and quantitative evaluation of treatment effects by considering more complex and subtle patterns. Keywords C. elegans, Machine learning, Computational ethology, Phenotypic screen, Drug screening, Disease models Currently, many diseases lack an approved treatment. This issue is even more pronounced in rare diseases, which, due to their low prevalence, receive less attention and fewer research resources1. Additionally, new diseases are discovered every year, further expanding this challenge2. To accelerate drug discovery, researchers turn to model organisms such as Caenorhabditis elegans (C. elegans) . This nematode has a fully sequenced, simple genome3, with 60-80% of its genes homologous to those in humans4. It can be genetically manipulated relatively easily using techniques such as RNA interference (RNAi) and CRISPR/Cas9 gene editing. Its small size (1 mm in length) and short lifespan (2–3 weeks) allow for large-scale experiments at a low cost. C. elegans has proven tobe a versatile and cost-effective tool for modeling diseases such as Alzheimer’s5, Parkinson’s6 and amyotrophic lateral sclerosis (ALS)7. Manual analysis of C. elegans behaviour for large-scale phenotypic screening is a laborious and timeconsuming process. For this reason, in recent years, video capture devices and algorithms have been developed that are capable of tracking and extracting features automatically8–14. Recently, a drug repurposing method was proposed for C. elegans models of Mendelian human diseases15. The authors use the CRISPR gene-editing technique to create a model of the disease they aim to study. Subsequently, screening assays are performed by treating the diseased worms with an FDA-approved drug library using a high-throughput imaging platform14. This platform captures high-resolution images and extracts a series of morphological, postural, and movement-related features13. Using block permutation t-tests with BenjaminiYekutieli correction for multiple comparisons, they determine how many features show statistically significant differences compared to the control strain. In an initial screening of 743 drugs with few replicates, they identify 30 potential hits, prioritizing those that shift three core features toward wild-type levels. In a second confirmation assay, more replicates are performed Instituto de Automática e Informática Industrial, Universitat Politècnica de València, Camino de Ver

FAQs

Why is automated behavioral phenotyping for Caenorhabditis elegans needed?

Manual analysis is laborious and time-consuming for large-scale phenotypic screening. Automated video capture and tracking enable feature extraction at higher throughput.

What problem do statistical feature-testing methods have in drug repurposing?

They rely on multiple-comparison corrections that can reduce statistical power, produce binary significance outcomes, and may be insufficient for detecting subtle non-linear interactions among features.

Which model performed best and how is it justified?

A Random Forest classifier trained on Tierpsy Tracker-extracted features achieved higher accuracy and explainability than deep neural networks that analyze video sequences directly, making it more suitable for drug testing experiments.

Aug 3, 2026

Upload by Ophelia