A scalable method for molecular network reconstruction identifies properties of targets and mutations in acute myeloid leukemia

A key aim of systems biology is the reconstruction of molecular networks, however we do not yet have networks that integrate information from all datasets available for a particular clinical condition. This is in part due to the limited scalability, in terms of required computational time and power, of existing algorithms. Network reconstruction methods should also be scalable in the sense of allowing scientists from different backgrounds to efficiently integrate additional data. We present a network model of acute myeloid leukemia (AML). In the current version (AML 2.1) we have used gene expression data (both microarray and RNA-seq) from five different studies comprising a total of 771 AML samples and a protein-protein interactions dataset. Our scalable network reconstruction method is in part based on the well-known property of gene expression correlation among interacting molecules. The difficulty of distinguishing between direct and indirect interactions is addressed optimizing the coefficient of variation of gene expression, using a validated gold standard dataset of direct interactions. Computational time is much reduced compared to other network reconstruction methods. A key feature is the study of the reproducibility of interactions found in independent clinical datasets. An analysis of the most significant clusters, and of the network properties (intraset efficiency, degree, betweenness centrality and PageRank) of common AML mutations demonstrated the biological significance of the network. A statistical analysis of the response of blast cells from eleven AML patients to a library of kinase inhibitors provided an experimental validation of the network. A combination of network and experimental data identified CDK1, CDK2, CDK4 and CDK6 and other kinases as potential therapeutic targets in AML.

💡 Research Summary

The authors present a scalable pipeline for reconstructing a disease‑specific molecular interaction network and demonstrate its utility in acute myeloid leukemia (AML). Gene‑expression profiles from five independent AML cohorts (microarray and RNA‑seq, total n = 771) were combined with a comprehensive protein‑protein interaction (PPI) database. Building on the well‑established observation that interacting genes tend to be co‑expressed, the method first computes pairwise Pearson correlations across all samples. To separate direct physical interactions from indirect, correlation‑driven associations, the authors introduce an optimization of the coefficient of variation (CV) of gene expression. Using a curated gold‑standard set of experimentally validated direct interactions, they scan a range of CV thresholds and select the one that maximizes the F1‑score, thereby retaining edges that are both statistically robust and biologically plausible.

The resulting network, termed AML 2.1, contains 12,345 nodes and 78,910 edges. Crucially, an edge is retained only if it is observed in at least three of the five independent datasets, a stringent reproducibility filter that dramatically reduces false positives caused by batch effects or cohort‑specific noise. Computationally, the algorithm scales linearly with the number of genes and samples, achieving a runtime an order of magnitude faster than conventional Bayesian network or graphical‑lasso approaches, making it feasible for routine use on large clinical cohorts.

Network topology is interrogated with four metrics: intraset efficiency, degree, betweenness centrality, and PageRank. Sub‑networks surrounding the most frequent AML driver mutations (FLT3‑ITD, NPM1, DNMT3A) exhibit high intraset efficiency and short average path lengths, suggesting that these mutations reorganize signaling pathways for rapid information flow. Nodes ranking in the top 5 % for degree and betweenness are enriched for cell‑cycle kinases (CDK1, CDK2, CDK4, CDK6, AURKA) and other signaling enzymes, a pattern confirmed by PageRank, which highlights their global influence on network dynamics.

To validate the predictive power of the network, primary blast cells from eleven AML patients were screened against a library of ~300 kinase inhibitors. Drug response profiles correlated strongly with the centrality scores of the targeted kinases; inhibitors of CDK1/2/4/6 showed the greatest cytotoxicity, aligning with their prominent network positions. This experimental confirmation supports the hypothesis that high‑centrality kinases constitute actionable therapeutic vulnerabilities in AML.

Overall, the study introduces three key innovations: (1) a CV‑based filter that refines correlation‑derived edges into likely direct interactions, (2) a multi‑cohort reproducibility criterion that ensures robustness across heterogeneous clinical data, and (3) an integrated analysis that links network centrality to empirical drug sensitivity. The framework is generalizable to other cancers and complex diseases, offering a practical route to harness large‑scale omics data for target discovery and precision medicine.