Jul 2026· npj Systems Biology and Applications· 0 citations
Medicine
TL;DR
The Wrapshap framework presents a novel FS method tailored for high-dimensional biological data, addressing challenges such as redundancy, multicollinearity, and interaction effects and opening new possibilities for real-time, interactive FS with user involvement.
Abstract
Among the existing feature selection (FS) methods in machine learning (ML), those known as wrapper methods often produce the most effective subset of features. However, their high computational cost and tendency to overfit make them impractical for high-dimensional data, as is frequently encountered in biological research. The Wrapshap framework presents a novel FS method tailored for high-dimensional biological data, addressing challenges such as redundancy, multicollinearity, and interaction effects. By leveraging the local accuracy property of Additive Local Explanation methods of SHAP, Wrapshap translates complex biological data into interpretable insights. The framework, integrating a trained ML model, allows explanations to be viewed as a new exploratory data space containing a comprehensive understanding of the underlying complexity of the raw data structure. Benchmarks against 84 regression and 106 classification datasets demonstrate its superiority in computational speed, feature ranking quality, and predictive performance compared to state-of-the-art methods like Recursive Feature Elimination. The method's adaptability extends to various ML models, explanation methods and tasks. By requiring only a single training of the complex model, Wrapshap mitigates the computational inefficiencies of traditional wrapper methods while maintaining performance monitoring thus opening new possibilities for real-time, interactive FS with user involvement. This innovative approach addresses a notable gap in existing methods, and promotes informed biological hypothesis generation, enhancing the interpretability and usability of ML in biomedical research.
Feature Selection (FS) is an essential data preprocessing technique aimed at identifying the most relevant features for predictive models. However, traditional FS approaches often struggle to capture complex interactions in real-world datasets, limiting their ability to fully support high-performing Machine Learning (M...
This study comprehensively benchmarked these methods for predicting the properties of small organic molecules and showed that TFM consistently outperform XGBoost, CatBoost, multilayer perceptrons, and other descriptor-based methods, even with careful hyperparameter selection for the baselines.
D. Matyushin, A. Sholokhova· Journal of Chemical Informat...· 0 citations
High-dimensional bioinformatics data, characterized by a large number of features relative to the number of samples, pose major challenges such as the ``curse of dimensionality,''leading to overfitting, high computational cost, and poor generalization. Traditional feature selection methods often suffer from limited sca...
Novel algorithms for estimating Lasso regression parameters by reformulating the problem as an inverse single-point optimization task are introduced, eliminating the need for explicit regularization parameter specification while maintaining robust feature selection capabilities and effective multicollinearity mitigatio...
In recent years, the issue of reproducibility in scientific studies has been at the forefront of scientific discussion. Many of these issues arise due to poor employment of statistical methods, making it crucial to address these knowledge gaps for frequently used methods. Elastic net regression (ENET) is one such algor...
Tristan A. Moxley, B. Ridenhour· bioRxiv· 0 citations
Feature selection plays a vital role in machine learning and data mining by identifying a representative subset to improve model performance, reduce computational cost, and prevent overfitting. Although traditional methods aim to eliminate redundant and irrelevant features, research into feature interactions has not be...
Shui-Fei Zheng, Bin Nie, Zhi-Yao Zhang et al.· PeerJ Computer Science· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.