Scientific Reports 5: Article number: 15147; published online: 20 October 2015; updated: 20 May 2016.
Introduction
This document serves as an official erratum for the 2015 study "Predictive analytics of environmental adaptability in multi-omic network models," originally published in Scientific Reports. The primary purpose of this update is to correct typographical errors found in the Results section, specifically under the subheading regarding the METRADE method. These corrections are vital for the accurate implementation and replication of the mathematical framework proposed by the authors.
The Challenge of Multi-omic Integration
A central challenge in systems biology is the integration of high-throughput "omic" data—such as transcriptomics and proteomics—into functional metabolic models. Traditional Flux Balance Analysis (FBA) often relies on Boolean logic, where genes are categorized as either "on" or "off" to constrain metabolic reactions. However, biological regulation is rarely binary; it is a continuous and nuanced process.
The original research introduced METRADE (MEtabolic TRanslation and ADaptation Enzyme-integrated model) to address this limitation. METRADE was designed to bridge the gap between gene expression levels and metabolic flux by utilizing continuous mapping functions, allowing for a more realistic representation of cellular metabolism.
The METRADE Framework
The METRADE method enhances standard FBA through three primary mechanisms:
- Continuous Mapping: It employs a function $h(\cdot)$ to map gene expression arrays directly to reaction bounds, moving beyond simple binary constraints.
- Bilevel Optimization: The model solves nested optimization problems, typically maximizing a biological objective like biomass production while simultaneously minimizing the discrepancy between predicted fluxes and experimental data.
- Variance-Weighted Integration: The model accounts for the reliability of the data by incorporating the variance of gene expression sets associated with specific metabolic reactions.
Mathematical Corrections
The erratum identifies and fixes critical typographical errors in the core equations defining the relationship between transcriptomic data and metabolic constraints. Specifically, the function $k(y_i)$ has been updated to its correct form:
$$k(y_i) = \left[ 1 + \frac{\gamma}{\sigma_i^2} | \log(y_i) | \right]^{\text{sgn}(y_i - 1)}$$
In this corrected formula:
- $y_i$ represents the magnitude of gene expression.
- $\sigma_i^2$ represents the variance of the gene set associated with a reaction.
- $\gamma$ is a weighting parameter.
- The inclusion of the absolute value of the log-expression and the explicit variance term ensures that the model correctly weights the "reliability" of the signal. By dividing by $\sigma_i^2$, the model prioritizes gene expression data that shows high consistency across replicates.
Implications for Systems Biology
These mathematical corrections reinforce the shift toward continuous regulation models in computational biology. By accurately formalizing the $k(y_i)$ function, the METRADE framework can more precisely predict how an organism's metabolism responds to subtle environmental or genotypic changes.
While the erratum does not present new experimental datasets, it provides the necessary mathematical rigor to support the predictive accuracy reported in the original 2015 study. For researchers utilizing multi-omic network models, these corrected formulas are essential for ensuring that computational results align with the intended biological logic of the METRADE algorithm.
Summary updated