Q1: Why should I use Graph wavelet based denoising ?

Answer: Noise and batch effect in single-cell expression profiles can obstruct the correct estimation of dependencies among genes and regulatory changes. GWNet uses graph-wavelet filters for improving gene-network based analysis of transcriptome. Our approach improves performance of several gene-network inference methods.

Q2: How is this different from KNN based smoothing/ imputation ?

Answer : The imputation methods predict the missing expression values in scRNA-seq profiles while our approach (GWNet) uses graph of cells made using KNN based approach and then filters the expression in transformed domain.

Q3: What is the format of input matrix provided to Graph Wavelet ?

Answer : Expression csv file should not contain header and genes i.e. it consist of only data on which filtering is going to perform. Row represents samples and column represent genes in csv file. Single cell data should be fpkm.

Q4:What python version and libraries are required for smoot running of the code ?

Answer : The python version required is 3.0+ and libraries required are: Numpy, Pandas, Pygsp, scipy, sklearn, networkx in python and minet,methods,GENIE3,propr in R