The pyNBS Algorithm

The following list is a step-by-step list of the key steps in the NBS algorithm. For additional details of each step, please refer to the function description of the linked module page. These steps closely reflect the individual steps described in the Jupyter Notebook walkthroughs of pyNBS.

1. Loading pyNBS required files

• Load somatic mutation data (see load_binary_mutation_data in data_import_tools)
• Load molecular network (see load_network_file in data_import_tools)

2. Pre-computing regularization network*

• Compute regularization network graph laplacian (see network_inf_KNN_glap in pyNBS_core)

3. Pre-computing network propagation influence matrix*

• Compute random-walk based influence of all genes to all other genes (see network_propagation in network_propagation)

4. Calculating multiple patient stratifications for consensus clustering

• Subsample binary somatic mutation data (see subsample_sm_mat in pyNBS_core)
• Propagate subsampled data over molecular network (see network_propagation or network_kernel_propagation in network_propagation)
• Quantile normalization of network-smoothed data (see qnorm in pyNBS_core)
• Network-regularized non-negative matrix factorization (netNMF) (see mixed_netNMF in pyNBS_core)

5. Consensus Clustering

• Construct consensus patient clusters (see mixed_netNMF in pyNBS_core)

6. pyNBS Clustering Plots

• Consensus co-clustering map (see plot_cc_map in pyNBS_plotting)
• Kaplan-Meier survival plots of clusters (see cluster_KMplot in pyNBS_plotting)