Biological network mapping and source signal deduction.

Area of Science:

• Systems Biology
• Bioinformatics
• Network Science

Background:

• Biological networks, such as transcriptional regulation and metabolism, are often bipartite.
• Understanding network architecture and source signals is crucial but often implicit in data.
• Existing methods may require prior information for network analysis.

Purpose of the Study:

• To develop a novel technique, Network Component Mapping (NCM), for deducing bipartite network connectivity and regulatory signals from data.
• To analyze complex biological data, including metabolite mixtures and gene expression, without prior network knowledge.

Main Methods:

• Network Component Mapping (NCM) algorithm to infer network structure and source signals.
• Application to UV-vis spectra of metabolite mixtures.
• Analysis of Saccharomyces cerevisiae gene expression data under different environmental conditions.

Main Results:

• NCM achieved higher resolution in deducing hidden mixing networks and pure component spectra compared to existing bipartite techniques.
• Inferred transcription networks from gene expression data were consistent with ChIP-chip data.
• NCM identified hidden transcription networks and combinatorial regulation in clean gene expression data.
• For noisy data, NCM provides the sparsest network consistent with observations.
• The method can incorporate partial network topology knowledge as constraints.

Conclusions:

• NCM is a powerful tool for uncovering hidden structures and signals in bipartite biological networks.
• The technique demonstrates utility in analyzing diverse biological datasets, from metabolite mixtures to gene expression.
• NCM offers a robust approach for network inference, adaptable to data quality and prior knowledge.