Optimal algorithm selection based on similarity among expression datasets and its potential to improve network-inference accuracy.

<p>Red lines show performance of Top<i>k</i>Net integrating algorithms that are optimal for a dataset with high-similarity, while green lines show that with low-similarity. Blue lines show performance of Top<i>k</i>Net integrating top 10 highest-performance algorithms. Dashed lines in red, green, and blue represent performance of community prediction integrating algorithms that are optimal for a dataset with high-similarity, that with low-similarity, and top 10 highest-performance algorithms, respectively. (<b>A</b>) AUC-PR for <i>E. coli</i> dataset. (<b>B</b>) AUC-ROC for <i>E. coli</i> dataset. (<b>C</b>) Max f-score for <i>E. coli</i> dataset. (<b>D</b>) AUC-PR for <i>S. cerevisiae</i> dataset. (<b>E</b>) AUC-ROC for <i>S. cerevisiae</i> dataset. (<b>F</b>) Max f-score for <i>S. cerevisiae</i> dataset.