variable_ids(result::FWResult{T}) -> Vector{T}

Extract the IDs/names of all variables (nodes) in the network.

Accelerated version of that found in MultipleTesting.jl

graph(result::FWResult{T}) -> SimpleWeightedGraph{Int, Float64}

Extract the underlying weighted graph from network results.

learn_network(all_data_paths::AbstractVector{<:AbstractString}, meta_data_path::AbstractString) -> FWResult{<:Integer}

Works like learnnetwork(datapath::AbstractString, metadatapath::AbstractString), but takes paths to multiple data sets (independent sequencing experiments (e.g. 16S + ITS) for the same biological samples) which are normalized independently.

learn_network(data_path::AbstractString, meta_data_path::AbstractString) -> FWResult{<:Integer}

Works like learn_network(data::AbstractArray{<:Real, 2}), but instead of a data matrix takes file paths to an OTU table and optionally a meta data table as an input.

• data_path - path to a file storing an OTU count matrix (and JLD2 meta data)

• meta_data_path - optional path to a file with meta data

• *_key - HDF5 keys to access data sets with OTU counts, Meta variables and variable names in a JLD2 file. If a data item is absent the corresponding key should be 'nothing'. See '?load_data' for additional information.

• verbose - print progress information

• transposed - if true, rows of data are variables and columns are samples

• kwargs... - additional keyword arguments passed to learn_network(data::AbstractArray{<:Real, 2})

learn_network(data::AbstractArray{<:Real, 2}) -> FWResult{<:Integer}

Learn an interaction network from a data matrix (including OTUs and optionally meta variables).

• data - data matrix with information on OTU counts and (optionally) meta variables

• header - names of variable columns in data

• meta_mask - true/false mask indicating which variables are meta variables

Algorithmic parameters

• heterogeneous - enable heterogeneous mode for multi-habitat or -protocol data with at least thousands of samples (FlashWeaveHE)

• sensitive - enable fine-grained association prediction (FlashWeave-S, FlashWeaveHE-S), sensitive=false results in the fast modes (FlashWeave-F, FlashWeaveHE-F)

• max_k - maximum size of conditioning sets, high values can lead to the removal of more spurious edgens, but may also strongly increase runtime and reduce statistical power. max_k=0 results in no conditioning (univariate mode)

• alpha - statistical significance threshold at which individual edges are accepted

• conv - convergence threshold, e.g. if conv=0.01 assume convergence if the number of edges increased by only 1% after 100% more runtime (checked in intervals)

• feed_forward - enable feed-forward heuristic

• fast_elim - enable fast-elimiation heuristic

• max_tests - maximum number of conditional tests that is performed on a variable pair before association is assumed

• hps - reliability criterion for statistical tests when sensitive=false

• FDR - perform False Discovery Rate correction (Benjamini-Hochberg method) on pairwise associations

• n_obs_min - don't compute associations between variables having less reliable samples (non-zero samples if heterogeneous=true) than this number. -1: automatically choose a threshold.

• time_limit - if feed-forward heuristic is active, determines the interval (seconds) at which neighborhood information is updated

General parameters

• normalize - automatically choose and perform data normalization method (based on sensitive and heterogeneous)

• track_rejections - store for each discarded edge, which variable set lead to its exclusion (can be memory intense for large networks)

• verbose - print progress information

• transposed - if true, rows of data are variables and columns are samples

• prec - precision in bits to use for calculations (16, 32, 64 or 128)

• make_sparse - use a sparse data representation (should be left at true in almost all cases)

• make_onehot - create one-hot encodings for meta data variables with more than two categories (should be left at true in almost all cases)

• update_interval - if verbose=true, determines the interval (seconds) at which network stat updates are printed

• extra_data - tuples of the form (data, header) representing counts from additional sequencing experiments (e.g. 16S + ITS) for the same biological samples. These will be normalized independently.

• share_data - if local parallel workers are detected, share input data (instead of copying)

• experimental_kwargs - experimental keyword arguments that are directly passed to the underlying inference engine

load_data(data_path::AbstractString, meta_path::AbstractString) -> (AbstractArray{<:Real, 2}, Vector{String}, AbstractArray{<:Real, 2}, Vector{String})

Load matrices with OTU count and optionally meta data from disc. Available formats are '.tsv', '.csv', '.biom' and '.jld2'.

• data_path - path to a file storing an OTU count matrix

• meta_data_path - optional path to a file with meta variable information

• *_key - HDF5 keys to access data sets with OTU counts, Meta variables and variable names in a JLD2 file. If a data item is absent the corresponding key should be 'nothing'. See '?load_data' for additional information.

• transposed - if true, rows of data are variables and columns are samples

load_network(net_path::AbstractString) -> FWResult{Int}

Load network results from disk. Available formats are '.edgelist', '.gml' and '.jld2'. For GML, only files with structure identical to save_network('network.gml') output can currently be loaded. FlashWeave parameters that were used for network inference are only available when loading from JLD2.

• net_path - path from which to load the network results

normalize_data(data::AbstractArray{<:Real, 2}) -> (AbstractArray{<:Real, 2}, Vector{String}, Vector{Bool}, Vector{Bool})

Normalize data using various forms of centered-logratio transformation and discretization. This should only be used manually when experimenting with different normalization techniques.

• data - data matrix with information on OTU counts and (optionally) meta variables

• header - names of variable-columns in data

• meta_mask - true/false mask indicating which variables are meta variables

• test_name - name of a FlashWeave-specific statistical test mode, the corresponding normalization method will be chosen automatically

• norm_mode - identifier of a valid normalization mode ('clr-adapt', 'clr-nonzero', 'clr-nonzero-binned', 'pres-abs', 'tss', 'tss-nonzero-binned')

• filter_data - whether to remove samples with no counts and variables with zero variation from data

• verbose - print progress information

• prec - precision in bits to use for calculations (16, 32, 64 or 128)

• make_sparse, make_onehot - see docstring for "learn_network(data::AbstractArray{<:Real, 2})"

parameters(result::FWResult{T}) -> Dict{Symbol, Any}

Extract the used parameters from network results.

save_network(net_path::AbstractString, net_result::FWResult) -> Void

Save network results to disk. Available formats are '.edgelist', '.gml' and '.jld2'.

• net_path - output path for the network

• net_result - network results object that should be saved

• detailed - save additional information, such as discarding sets, if available (output file suffixes: 'rejections', 'unchecked')

take!(rr::RemoteChannel, args...)

Fetch value(s) from a RemoteChannel rr, removing the value(s) in the process.

Can't be used for MiTestCond since levels_z requires contingency table