Supporting Information
A Novel Network Profiling Analysis Reveals System Changes in Epithelial-Mesenchymal Transition
PLoS ONE 6(6): e20804. doi:10.1371/journal.pone.0020804

Teppei Shimamura^1,*, Seiya Imoto¹, Yukako Shimada², Yasuyuki Hosono², Atsushi Niida¹,
Masao Nagasaki¹, Rui Yamaguchi¹, Takashi Takahashi² and Satoru Miyano^1
1 Human Genome Center, Institute of Medical Science, University of Tokyo
² Nagoya University Graduate School of Medicine
^* Corresponding author

Correction to: PLoS One. 2011; 6(6): e20804. Epub 2011 Jun 7

We announced a mis-statement in our article.

In the section titled "Identification of relationships between regulators and epithelial-mesenchymal transition-related functional gene sets", the sentence:

"Since KLF4 increases the migration of cells [29] and induces EMT [10], these miRNAs might be related with EMT-related functions or control EMT by targeting KLF4."

should read:

"Since KLF4 decreases the migration of cells and loss of KLF4 induces EMT [10], these miRNAs might be related with EMT-related functions or control EMT by targeting KLF4."

We would like to thank Dr. Jennifer Yori for pointing out this mis-statement.

Supplementary Files

Supplements to Results:


• File S1 (compressed file (.tgz)):
  Edge lists of EMT-related modulator-dependent gene networks for 254 cancer cell lines (1^st - 254^th). The compressed file includes 254 comma-delimited text files. In each text file, the 1^st, 2^nd, and 3^rd columns represent parent gene, child gene, and the estimated coefficient between the two genes by NetworkProfiler, respectively.
• File S2 (compressed file (.tgz)):
  Edge lists of EMT-related modulator-dependent gene networks for 254 cancer cell lines (255^th - 508^th). The compressed file includes 254 comma-delimited text files. In each text file, the 1^st, 2^nd, and 3^rd columns represent parent gene, child gene, and the estimated coefficient between the two genes by NetworkProfiler, respectively.
• File S3 (compressed file (.tgz)):
  Edge lists of EMT-related modulator-dependent gene networks for 254 cancer cell lines (509^th - 762^th). The compressed file includes 254 comma-delimited text files. In each text file, the 1^st, 2^nd, and 3^rd columns represent parent gene, child gene, and the estimated coefficient between the two genes by NetworkProfiler, respectively.
• File S4 (compressed file (.tgz)):
  Enrichment scores between 1732 regulators and 5 functions for 762 cancer cell lines. The compressed file includes 762 tab-delimited text files. In each text file, the row and column indicate regulator and functional gene set, respectively. The (i,j)-th element represents the statistical significance (-log₁₀(q-value)) for the enrichment of target genes of the i-th regulator on the j-th function by using Fisher's exact test.

Supplements to Materials and Methods:


• File S5 (pdf file):
  Calculation of modulator mode of action with respect to the relationship between a regulator and its target gene.

Supplementary Figures

Supplements to Results:


• Figure S1:
  Quantitative real-time RT-PCR analysis of KLF5 in siKLF5-treated A549 cells.
• Figure S2:
  Expression profiles of miR-100 in order of ascending the EMT-related modulator values.
• Figure S3:
  miR-100-induced changes in biologic characteristics in NCI-H1437 and NCIH727 NSCLC cell lines. (a) Representative phase contrast microscopic images showing negligible changes in morphology by miR-100 introduction in both NSCLC cells lines. NC#2, negative control. (b) Motility assay showing increased migration by introduction of miR-100 in both NSCLC cell lines. *, P < 0.05.

Supplements to Materials and Methods:


• Figure S4:
  Histogram of computational times for inferring cancer cell line-specific gene networks running on 12 core CPUs. The 762 cancer cell line-specific gene networks related with the EMT were calculated from 13,508 × 762 gene expression data when 100 target genes were randomly selected among 13,508 genes and the number of regulators was not restricted, i.e., 1,732 regulators were used. The comptational times were based on 12 core CPUs (Intel Xeon Processor E5450 (# of cores = 4, clock speed = 3.0 GHz) × 3). The histogram was calculated by 100,000 iterations.
• Figure S5:
  Example of paths among four genes, R₁, T₂, R₃, and R₄.

Supplementary Tables

Supplements to Results:


• Table S1
  List of candidate regulators mapped to 1183 transcription factors and 47 nuclear receptors.
• Table S2
  List of candidate regulators mapped to 502 human microRNAs.
• Table S3
  List of coherent genes (p-value<10^-5) related to epithelial-mesenchymal transition calculated by extraction of expression module (EEM).
• Table S4
  EMT-related modulator values of 762 cancer cell lines calculated by signature-based hidden modulator extraction.
• Table S5
  List of 370 putative master regulators of E-cadherin during the EMT which were estimated by NetworkProfiler.
• Table S6
  List of 627 putative master regulators of E-cadherin which were estimated by a structual equation model (SEM) with the elastic net.
• Table S7
  Regulator function matrix between 1732 regulators and 5 functions. The row and column indicate regulator and functional gene set, respectively. The (i,j)-th element represents the change during the EMT in the statistical significance (-log₁₀(q-value)) for the enrichment of target genes of the i-th regulator on the j-th function. The last column indicate the integral q-value of each row regulator which were used to determine which regulator strongly affected the functional gene sets.
• Table S8
  List of 17 putative master regulators (integral q-value<10^-10) which correlated at least one or more EMT-related functions and were known to be downstream targets of TGFB1 with published evidence from Ingenuity Knowledge Base (http://www.ingenuity.com).
• Table S9
  List of the changes in the regulatory effects from 1732 regulators to E-cadherin and vimentin during the EMT.