network_benchmark_noise_filtering.RdThis function takes four types of matrices as input which were generated with the same network inference tool (e.g. GENIE3), but had different input data to the network inference tool in the context of noise-filtering and quantile-normalization. It is assumed that the four matrices inserted into this function come from network inference runs with the following input data specifications:
Raw mapped RNAseq count data without noise-filtering or quantile-normalization applied (adj_mat_not_filtered_not_normalized)
Raw mapped RNAseq count data with noise-filtering applied but no quantile-normalization (adj_mat_filtered_and_not_normalized)
Raw mapped RNAseq count data without noise-filtering but with quantile-normalization applied (adj_mat_not_filtered_but_normalized)
Raw mapped RNAseq count data with noise-filtering applied and with quantile-normalization applied (adj_mat_filtered_and_normalized)
All relevant pairwise comparisons are then performed internally with network_dist_pairwise_genes based on gene-wise
Hamming Distance or Jaccard Similarity Coefficient computations.
network_benchmark_noise_filtering( adj_mat_not_filtered_not_normalized, adj_mat_filtered_and_not_normalized, adj_mat_not_filtered_but_normalized, adj_mat_filtered_and_normalized, grn_tool = NA, threshold = "median", dist_type = "hamming", print_message = TRUE )
| adj_mat_not_filtered_not_normalized | a weighted adjacency matrix derived from a network inference program where no noise-filtering or quantile-normalization was applied to the input data. |
|---|---|
| adj_mat_filtered_and_not_normalized | a weighted adjacency matrix derived from a network inference program where noise-filtering but no quantile-normalization was applied to the input data. |
| adj_mat_not_filtered_but_normalized | a weighted adjacency matrix derived from a network inference program where no noise-filtering but quantile-normalization was applied to the input data. |
| adj_mat_filtered_and_normalized | a weighted adjacency matrix derived from a network inference program where noise-filtering and quantile-normalization were applied to the input data. |
| grn_tool | a character string specifying the gene regulatory network inference tool that was used to
generate input matrices. Default is |
| threshold | we recommended to use
See |
| dist_type | a distance method that shall be applied on the binary values for each gene. Available options are:
See |
| print_message | shall messages be printed? |
Hajk-Georg Drost
# Benchmark GENIE3 inferred networks with raw, no_noise, and quantile_norm combinations genie3_49_raw <- as.matrix(read.csv( system.file("data/network_raw_49_placenta_development.csv", package = "edgynode"), row.names = 1)) genie3_49_noNoiseCM_raw <- as.matrix(read.csv( system.file("data/network_noNoiseCM_raw_49_placenta_development.csv", package = "edgynode"), row.names = 1)) genie3_49_qnorm_no_noise_removed <- as.matrix(read.csv( system.file("data/network_qnorm_49_placenta_development.csv", package = "edgynode"), row.names = 1)) genie3_49_noNoiseCM_qnorm <- as.matrix(read.csv( system.file("data/network_noNoiseCM_qnorm_49_placenta_development.csv", package = "edgynode"), row.names = 1)) # Run Benchmark using Hamming distance benchmark_hamming <- network_benchmark_noise_filtering( genie3_49_raw, genie3_49_noNoiseCM_raw, genie3_49_qnorm_no_noise_removed, genie3_49_noNoiseCM_qnorm, dist_type = "hamming", grn_tool = "GENIE3")#> Warning: The matrix provided as input for network_rescale() was coerced into symmetric.#>#> Warning: The matrix provided as input for network_rescale() was coerced into symmetric.#>#> Warning: The matrix provided as input for network_rescale() was coerced into symmetric.#>#> Warning: The matrix provided as input for network_rescale() was coerced into symmetric.#>#>#>#>#>#>#>#>#>#>#>#>#>#>#>#>#>#>#>#>#>#>#>#>#>#>#># look at results benchmark_hamming#> # A tibble: 46 x 8 #> grn_tool genes `-F -N / +F -N` `-F -N / -F +N` `-F -N / +F +N` #> <chr> <chr> <dbl> <dbl> <dbl> #> 1 GENIE3 ENSM… 2 4 7 #> 2 GENIE3 ENSM… 10 11 14 #> 3 GENIE3 ENSM… 1 6 9 #> 4 GENIE3 ENSM… 3 4 6 #> 5 GENIE3 ENSM… 2 6 7 #> 6 GENIE3 ENSM… 2 5 6 #> 7 GENIE3 ENSM… 7 10 14 #> 8 GENIE3 ENSM… 1 8 9 #> 9 GENIE3 ENSM… 4 5 4 #> 10 GENIE3 ENSM… 4 1 1 #> # … with 36 more rows, and 3 more variables: `-F +N / +F +N` <dbl>, `-F +N / +F #> # -N` <dbl>, `+F +N / +F -N` <dbl># Run Benchmark using Jaccard Coefficients benchmark_jaccard <- network_benchmark_noise_filtering( genie3_49_raw, genie3_49_noNoiseCM_raw, genie3_49_qnorm_no_noise_removed, genie3_49_noNoiseCM_qnorm, dist_type = "jaccard", grn_tool = "GENIE3")#> Warning: The matrix provided as input for network_rescale() was coerced into symmetric.#>#> Warning: The matrix provided as input for network_rescale() was coerced into symmetric.#>#> Warning: The matrix provided as input for network_rescale() was coerced into symmetric.#>#> Warning: The matrix provided as input for network_rescale() was coerced into symmetric.#>#>#>#>#>#>#>#>#>#>#>#>#>#>#>#>#>#>#>#>#>#>#>#>#>#>#># look at results benchmark_jaccard#> # A tibble: 46 x 8 #> grn_tool genes `-F -N / +F -N` `-F -N / -F +N` `-F -N / +F +N` #> <chr> <chr> <dbl> <dbl> <dbl> #> 1 GENIE3 ENSM… 0.92 0.84 0.741 #> 2 GENIE3 ENSM… 0.643 0.577 0.5 #> 3 GENIE3 ENSM… 0.955 0.76 0.625 #> 4 GENIE3 ENSM… 0.842 0.810 0.714 #> 5 GENIE3 ENSM… 0.905 0.75 0.72 #> 6 GENIE3 ENSM… 0.929 0.833 0.793 #> 7 GENIE3 ENSM… 0.75 0.630 0.481 #> 8 GENIE3 ENSM… 0.96 0.704 0.667 #> 9 GENIE3 ENSM… 0.846 0.808 0.852 #> 10 GENIE3 ENSM… 0.818 0.947 0.947 #> # … with 36 more rows, and 3 more variables: `-F +N / +F +N` <dbl>, `-F +N / +F #> # -N` <dbl>, `+F +N / +F -N` <dbl>