Learning Center
Plans & pricing Sign in
Sign Out



									Considering more high-throughput interactome or transcriptome data will be available
in the near future, we propose a new pipeline for reconstructing the NP network by
integrating more datasets. The proposed pipeline involves the following three steps
(Figure S1).

Step 1. Different Arabidopsis PPI data are merged into a comprehensive interactome
by using network analysis tools (e.g., Advance Network Merge, a cytoscape plug-in).

Step 2. The PCC values for each pair of interacting proteins in different microarray
datasets are calculated. Then, a combined PCC can be deduced from the individual
PCC values and will be further employed to evaluate the correlation of gene
expression profiles. Here we show an example to obtain a combined PCC based on a
fixed effects model (Hedges, L. V. & Vevea, J. L. (1998). Fixed- and random-effects
models in meta analysis. Psychological Methods, 3: 486-504.) The combined PCC
value (PCCcombined) can be calculated from the following two equations.

                          1                              
                           log e  1  PCCi    si  3

                       2  1  PCC  
                                              
                    i 1                  i            
               Zr                   n                                             (S1)
                                     si  3
                                     i 1

                                        e 2 Zr 1
                           PCCcombined  2 Z 1                                  (S2)
                                        e r
where PCCi means the PCC value derived from the ith microarray dataset, si denotes
the number of samples in the ith microarray dataset, and n stands for the total number
of microarray datasets. To reconstruct the NP network, we can filter PPIs using the
calculated PCCcombined values (i.e., only correlated or anti-correlated PPIs is kept for
the reconstruction of the NP network).

Step 3. To infer network modules from the reconstructed NP network, each
microarray dataset should be centralized and normalized individually. Then, all
microarray data associated with the proteins in the reconstructed NP network can be
joined into a new microarray dataset. Finally, the new dataset is clustered and scanned
from the top of the hierarchical tree to obtain groups of genes (network modules) that
have less than 5% intra-group anti-correlated interactions.

Figure S1. A new pipeline for reconstructing the NP network by integrating more
interactome and transcriptome data.


To top