network_file_name = "network.tsv"
The network.tsv is a long-format TSV file containing Regulator -> Target edges. This TSV file is sorted by the confidence score of the regulator (TF) -> target (gene) edge, from largest to smallest.:
target regulator combined_confidences gold_standard precision recall MCC F1 BSU24750 BSU04730 0.999986 1 1 0.00165 0.04057 0.003295 BSU13020 BSU04730 0.999984 BSU09690 BSU04730 0.99998 BSU06590 BSU04730 0.999978 BSU18510 BSU04730 0.999976 BSU25800 BSU25810 0.999975
If the gene and TF are in the gold standard, the gold standard for this edge is reported (1 if present, 0 if not present), and the model performance is calculated. The Precision, Recall, MCC, and F1 scores are calculated assuming that all edges above a row (with greater confidence scores) are predicted TF -> Gene interactions, and all values below are predicted to not be TF -> Gene interactions. Rows which do not contain any gold standard (either 1 or 0) indicate that the regulator or the target are not in the Genes x TFs gold standard matrix. These rows will not be scored.
Also included is a column indicating if the network edge was in the prior (1, 0, or not present if the gene or TF were not
present in the prior network).
beta.sign.sum column is the number of times the model coefficient occurred and the sign
(positive model coefficients will be reported as a positive value, and negative model coefficients will be reported as a
var.exp.median column reports the median amount of variance in the gene explained by the regulator.
InferelatorResults(network_data, betas_stack, combined_confidences, metric_object, betas_sign=None, betas=None)¶
For network analysis, the results produced in the output_dir are sufficient. Model development and comparisons may require to values that are not written to files. An InferelatorResults object is returned by the
workflow.run()methods (A list of InferelatorResults objects is returned by the
This object allows access to most of the internal values created by the inferelator.
Results name, usually set to task name. Defaults to None.
Network dataframe, usually written to network.tsv.gz
The aggregate sign of non-zero betas. This is a dataframe which is Genes x TFs
Count of non-zero betas. This is a dataframe which is Genes x TFs
Confidence scores for tf-gene network edges. This is a dataframe which is Genes x TFs
Task result objects if there were multiple tasks. None if there were not. This is a dict, keyed by task ID