Our Team
Developed By:
Susara Ouchithya
Sahan Madhushanka
Kaleelur Rahman
Supervised By:
Dr. Nilmini Hettiarachchi (Tokyo Metropolitan Institute of Medical Science, Japan)
Dr. Dilhari Attygalle (University of Colombo, Department of Statistics, Sri Lanka)
Dr. Gayan Dharmarathne (University of Colombo, Department of Statistics, Sri Lanka)
Our Mission
For nearly a century, the scientific community has grappled with the challenge of Multiple Hypothesis Testing, particularly in gene expression analysis, where testing numerous hypotheses simultaneously can lead to a high risk of false positives. Despite extensive discussions and proposed solutions, a comprehensive framework has been elusive—until now.
Our Solution
Introducing MultiDST, a groundbreaking solution designed to enhance and unify existing procedures into a robust framework for multiple testing. Our Python library and user-friendly website streamline the process, allowing researchers to input their p-values and apply various correction methods to control both family-wise error rate (FWER) and false discovery rate (FDR). MultiDST supports multiple correction techniques, including Bonferroni, Holm, Sequential Goodness of Fit (SGoF), Benjamini-Hochberg, Benjamini-Yekutieli, and Storey’s Q value procedures, all validated through comprehensive simulation studies and power analysis.
Uniqueness
MultiDST stands out as the first platform to systematically compare six correction methods—Bonferroni, Holm, SGoF, Benjamini-Hochberg, Benjamini-Yekutieli, and Storey’s Q value—accessible through an intuitive interface. We introduce the Significant Index Plot (SIP), an innovative visualization tool that aids researchers in detecting significant hypotheses across these correction procedures. The SIP provides a clear representation of significant hypotheses under each method, simplifying interpretation and comparison of results. Additionally, our hybrid technique sequentially tests and removes hypotheses, allowing seamless navigation between FWER and FDR methods, ensuring optimal results across diverse data scenarios, including Independent Hypotheses Weighting (IHW) situations.
Proof of Concept
MultiDST has demonstrated significant potential through pilot tests with several research groups focusing on gene expression analysis. Researchers found that our hybrid technique and SIP provided clear, intuitive visualizations, significantly enhancing the interpretation of significant hypotheses. The user-friendly interface and robust analytical tools streamlined workflows, enabling more accurate and reliable findings.
Quality and Application of Technology
MultiDST exemplifies high-quality technological integration, combining advanced statistical methods with modern web development. The front end, built with Next.js, ensures a fast, scalable, and responsive interface, while the back end, supported by our Python package MultiDST (available on [PyPI] (https://pypi.org/project/multidst/)), manages API requests efficiently using Python's Fast-API. Researchers can easily input their p-values and interact with the platform seamlessly.