Scientific Reports · Nov 4, 2026

Integrated computational pipelines using machine learning and deep neural networks for modelling Drosophila cardiac aging and disease classification

By Kondapaneni V, Melkani Y, Melkani GC

High-speed optical microscopy of Drosophila hearts has become an essential approach for modelling cardiovascular disease and aging. However, manual evaluation of cardiac function is time-consuming, subjective, and prone to error, highlighting the need for reliable computational frameworks. Here, we present computational pipelines that combine segmentation-derived cardiac parameters with ensemble machine learning and raw video-based representations with deep convolutional neural networks, enabling the development of new predictive models while extending our previous framework. This integrated approach enables automated and quantitative phenotyping of cardiac function across both aging and cardiovascular disease contexts. We evaluated the performance of the pipelines across three biological problems relevant to cardiovascular systems biology: distinguishing between young (1-week-old) and aged (5-week-old) Drosophila hearts, classifying healthy hearts from those carrying the lamin C mutation (LamCG489V), and identifying phenotypes associated with Ogdh knockdown. Ensemble learning with Logistic Regression, Support Vector Machine, and XGBoost (age classification) and Random Forest (LamCG489V and Ogdh classification) applied to segmentation-derived features achieved accuracies (mean ± standard deviation across 5-fold cross-validation) of 83.7% ± 5.2% for aging classification, 78.2% ± 8.2% for LamCG489V mutant detection, and 98.5% ± 3.1% for Ogdh knockdown identification. In parallel, video-based modelling using a custom dual-input convolutional neural network and 3D residual networks (R3D-18 and MC3-18) achieved accuracies of 89.5% ± 4.3% for age classification, 75.6% for lamin C mutant detection, and 90.8% ± 6.0% for Ogdh knockdown classification. By integrating classical feature-based machine learning with modern deep learning architectures, this work provides a scalable computational framework for modeling Drosophila cardiac aging and disease. More broadly, the study demonstrates how multimodal machine learning can enhance quantitative phenotyping, reduce human error in analysis, and accelerate the application of models in systems-level investigations of cardiovascular health and aging.

doi.org/10.1038/s41598-026-75107-wOpen accessBiomarkers

AgingResearch.News