[Paper Review] BioStatFlow -Statistical Analysis Workflow for "Omics" Data

Daniel Jacob, Catherine Deborde+1 more|arXiv (Cornell University)|Jul 9, 2020

Metabolomics and Mass Spectrometry Studies7 references4 citations

TL;DR

BioStatFlow is a free, user-friendly web application that enables biologists without advanced statistical expertise to perform end-to-end 'omics' data analysis using R-based workflows. It guides users through normalization, imputation, univariate and multivariate analyses, and network visualization, with outputs easily saved or downloaded.

ABSTRACT

BioStatFlow is a free web application, useful to facilitate the performance of statistical analyses of "omics", including metabolomics, data using R packages. It is a fast and easy on-line tool for biologists who are not experts in univariate and multivariate statistics, do not have time to learn R language, and only have basic notions in biostatistics. It guides the biologist through the different steps of a statistical workflow, from data normalization and imputation of missing data to univariate and multivariate analyses. It also includes tools to reconstruct and visualize networks based on correlations. All outputs are easily saved in a session or downloaded. New analytical modules can be easily included upon request. BioStatFlow is available online: http://biostatflow.org

Motivation & Objective

• Address the challenge of statistical analysis in 'omics' data for biologists lacking advanced statistical training.
• Provide a streamlined, interactive workflow to guide non-expert users through complex statistical procedures.
• Reduce the learning curve associated with R programming and multivariate statistics in metabolomics and similar fields.
• Facilitate reproducible and shareable analysis sessions with easy export and save functionality.
• Support extensibility by allowing new analytical modules to be added upon request.

Proposed method

• Leverages existing R packages to perform statistical operations within a web-based interface.
• Implements a step-by-step workflow guiding users from data import to final visualization.
• Automates data normalization and missing value imputation using established statistical methods.
• Integrates univariate tests (e.g., t-tests, ANOVA) and multivariate techniques (e.g., PCA, PLS-DA) for 'omics' data.
• Enables correlation-based network reconstruction and visualization for biological insight.
• Stores analysis sessions and allows export of results in multiple formats for reproducibility.

Experimental results

Research questions

• RQ1 How can non-expert biologists perform robust statistical analysis on 'omics' data without learning R?
• RQ2 What workflow design maximizes usability while maintaining statistical rigor for metabolomics and similar 'omics' fields?
• RQ3 Can a web-based interface effectively encapsulate complex R-based statistical pipelines for non-programmers?
• RQ4 How can reproducibility and data sharing be enhanced in biological data analysis for non-expert users?
• RQ5 To what extent can extensible modules be integrated into a unified analytical platform for 'omics' data?

Key findings

• BioStatFlow successfully enables biologists with minimal statistical training to perform comprehensive 'omics' data analysis through a guided web interface.
• The application supports key analytical steps including data normalization, imputation, univariate and multivariate testing, and network visualization.
• All outputs are persistently saved within the session or downloadable, supporting reproducibility and collaboration.
• The platform is extensible, allowing new analytical modules to be added upon request, enhancing long-term utility.
• The tool significantly reduces the barrier to entry for statistical analysis in 'omics' research by abstracting complex R code into an intuitive interface.
• The application is publicly available at the provided URL, ensuring broad accessibility for researchers in the life sciences.