iNAP

Integrated Network Analysis Pipeline (https://inap.denglab.org.cn/)

Notice: The iNAP website is updated and please make sure you are visiting our new iNAP website.

Please see more details in the iMeta paper and Chinese Version:

Kai Feng, Xi Peng, Zheng Zhang, Songsong Gu, Qing He, Wenli Shen, Zhujun Wang, Danrui Wang, Qiulong Hu, Yan Li, Shang Wang, Ye Deng. iNAP: an integrated Network Analysis Pipeline for microbiome studies. iMeta. 2022,1:e13. https://doi.org/10.1002/imt2.13

iNAP procedure file: https://inap.denglab.org.cn/static/iNAP-Denglab-Jan.2022.pdf

iNAP tutorial video: in Youtube or Bilibili.

Network visualization video: in Youtube or Bilibili.

This is a descriptive file to run some network calculations for iNAP, including SparCC, SPIEC-EASI and eLSA/LA. The codes here are mainly dumped from the original workflow of each method and modified minorly for iNAP.

SparCC

Details for SparCC can be found at dlegor/SparCC, which provides a SparCC script for python3 version. The installation and relevant introduction is available as well.

Here we only introduce brief script to obtain SparCC results after the Step of Majority selection in iNAP.

Plant_6_microbe_8.txt is the downloaded file from iNAP after majority selection. The default values can be found in iNAP.

SparCC.py Plant_6_microbe_8.txt  -i 20 -x 10 -t 0.1 -c SparCC_correlation.txt 
mkdir Plant_6_microbe_8
MakeBootstraps.py Plant_6_microbe_8.txt -n 100 -t permutation_#.txt -p Plant_6_microbe_8/ 
for i in `seq 0 99`; 
do 
SparCC.py Plant_6_microbe_8/permutation_$i.txt -i 20 -x 10 -t 0.1 -c Plant_6_microbe_8/perm_cor_$i.txt  
done
PseudoPvals.py SparCC_correlation.txt Plant_6_microbe_8/perm_cor_#.txt 100 -o SparCC_pseudo_p_value.txt -t two_sided 

• 20: Number of inference iteration to average over
• 10: Number of exclusion iterations to remove strongly correlated pairs
• 0.1: Correlation strengh exclusion threshold
• 100: Number of shuffled times
• two_sided: one_sided or two_sided

SPIEC-EASI

Details for SPIEC-EASI can be found at SPIEC-EASI, which provides detailed installation and relevant introduction.

Here we only introduce brief script to obtain SPIEC-EASI results of interdomain interactions in iNAP.

use 6 for plant abundance data and 8 for microbial data as majority selection criterion.

library(SpiecEasi)
majority_otu <- 8
majority_plant <- 6
otu_table <- read.table("otu.txt",header = T, row.names = 1,sep = "\t")
plant_table <- read.table("plant.txt",header = T, row.names = 1,sep = "\t")
samp.name <- interaction(colnames(otu_table),colnames(plant_table))
otu_table[is.na(otu_table)] = 0
otu_table2<-otu_table[,samp.name] 
otu_table2[otu_table2>0] <- 1
otu_table_MV <- otu_table[rowSums(otu_table2)>= majority_otu,]
plant_table[is.na(plant_table)] = 0
plant_table2<-plant_table[,samp.name] 
plant_table2[plant_table2>0] <- 1
plant_table_MV <- plant_table[rowSums(plant_table2)>= majority_plant,]
rename_otu<-function(x,y){
result<-paste(x,"_",y,sep = "")
}
rownames(otu_table_MV)<-sapply(rownames(otu_table_MV),rename_otu,y="M")
rownames(plant_table_MV)<-sapply(rownames(plant_table_MV),rename_otu,y="P")
sp.res <- multi.spiec.easi(datalist=list(t(otu_table_MV),t(plant_table_MV)),method="glasso",sel.criterion = "stars",verbose=FALSE,pulsar.select = TRUE,lambda.min.ratio=0.1,nlambda=50,pulsar.params = list(rep.num=50,thresh=0.05,ncores=8),cov.output=TRUE)
opticov <- getOptiCov(sp.res)
colnames(opticov) = rownames(opticov) <- colnames(sp.res[["est"]][["data"]])
write.table(as.matrix(opticov),file = "SpiecEasi_result.txt",sep = "\t",quote=FALSE,col.names=NA)

eLSA/LA

Details for eLSA/LA can be found at elsa, which provides detailed installation and relevant introduction.

Here we only introduce brief script to obtain LSA results after the Step of Majority selection in iNAP.

lsa_compute Plant_6_microbe_8.txt LSA_calculation_result.txt -s 4 -r 1 -d 3 -p perm -x 100 -b 0 -f none -t simple -n robustZ -q scipy 

• 4: number of spots
• 1: number of replicates for each time spot
• 3: maximum time delay
• perm: Use permutation; theo Theoretical approximation; mix Use theoretical approximation for pre-screening and then use permutation;
• 100: Permutation number 100 or precision=0.01/permutation for p-value estimation
• 0: Number of bootstraps for 95% confidence interval estimation
• none: fill up with zeros; zero fill with zero order splines; linear fill up with linear splines;quadratic fill up with quadratic spline;cubic fill up with cubic spline; slinear fill up with slinear; nearest fill up with nearest neighbor
• simple: simple averaging; SD standard deviation weighted averaging; Med simple Median; MAD median absolute deviation weighted median;
• robustZ: percentileZ normalization + robust estimates (with perm, mix and theo, and must use this for theo and mix, default); percentile percentile normalization, including zeros (only with perm); percentileZ percentile normalization + Z-normalization; pnz percentile normalization, excluding zeros (only with perm); none none;
• scipy: Qvalue calculation method

Pearson/Spearman correlations

Here we provided one alternative correlation calculation scripts for Pearson/Spearman methods using 'WGCNA' package in R program.

This script would get similar correlation matrix as in iNAP pipeline with 'keep blank' for missing values, and the P values were adjusted using 'fdr' methods, which depends on you. The P values in iNAP were not adjusted, but you can using similar methods as shown here to adjust.

library(WGCNA)
otu <- read.table("Plant_6_microbe_8.txt",header = T,row.names = 1,sep="\t")
corres <- corAndPvalue(t(otu),use="pairwise.complete.obs",method = c("spearman"))
corres.r <- corres$cor
corres.p <- corres$p
corres.p.adj <- p.adjust(corres.p[upper.tri(corres.p,diag = F)],method = "fdr")
corres.p.adj.matrix <- matrix(0,nrow=nrow(corres.p),ncol=ncol(corres.p))
corres.p.adj.matrix[lower.tri(corres.p,diag = F)] <- corres.p.adj
corres.p.adj.matrix<-as.matrix(as.dist(corres.p.adj.matrix))
colnames(corres.p.adj.matrix) = rownames(corres.p.adj.matrix) <- rownames(corres.p)
write.table(corres.r,"Correlation_matrix_r_values.txt",sep="\t",col.names = NA)
write.table(corres.p.adj.matrix,"Correlation_matrix_p_adj_values.txt",sep="\t",col.names = NA)

• spearman: using "pearson" or "spearman"
• fdr: using "fdr", "bonferroni", "holm", "hochberg", "hommel", "BH", "BY" or other adjustment methods