Reed-CompBio · tristan-f-r · Oct 8, 2025 · Oct 11, 2025 · Oct 11, 2025 · Oct 11, 2025
diff --git a/docs/prms/allpairs.rst b/docs/prms/allpairs.rst
@@ -8,3 +8,17 @@ with respect to a cartesian product.
 
 All Pairs does not take any arguments. Its source code is at https://github.com/Reed-CompBio/all-pairs-shortest-paths
 licensed under MIT.
+
+Dataset Usage
+-------------
+
+All Pairs Shortest Paths uses ``sources``, ``targets``, and edge weights.
+All Pairs Shortest Paths uses edge direction information.
+However, All Pairs Shortest Paths always returns an undirected subnetwork.
+
+Implementation Details
+----------------------
+
+Internally, All Pairs Shortest Paths only accepts fully undirected or directed graphs.
+SPRAS will automatically convert edges to directed edges as necessary.
+For more information, see the section on :ref:`algorithm directionality <directionality>`.
diff --git a/docs/prms/bowtiebuilder.rst b/docs/prms/bowtiebuilder.rst
@@ -1,5 +1,5 @@
 BowTieBuilder
-==========
+=============
 
 BowTieBuilder is a pathway reconstruction algorithm which constructs pathways in a 'bowtie'-like
 fashion, finding the intersections of shortest paths between sources and targets and using those nodes as a basis
@@ -9,3 +9,15 @@ BowTieBuilder does not take in any arguments.
 
 * Repository: https://github.com/Reed-CompBio/BowTieBuilder-Algorithm
 * Paper: https://doi.org/10.1186/1752-0509-3-67
+
+Dataset Usage
+-------------
+
+BowTieBuilder uses ``sources``, ``targets``, edge weights, and edge direction information.
+
+Implementation Details
+----------------------
+
+BowTieBuilder's internal implementation only takes in directed interactomes.
+SPRAS will automatically convert edges to directed edges as necessary.
+For more information, see the section on :ref:`algorithm directionality <directionality>`.
diff --git a/docs/prms/domino.rst b/docs/prms/domino.rst
@@ -9,8 +9,15 @@ DOMINO has two optional parameters:
 * slice_threshold: the p-value threshold for considering a slice as relevant
 * module_threshold: the p-value threshold for considering a putative module as final module
 
-Wrapper Details
-===============
+Dataset Usage
+-------------
+
+DOMINO requires the `active` column to be set. DOMINO does not consider edge weights,
+nor graph directionality: directed edges are treated as undirected edges, and DOMINO
+returns an undirected subnetwork.
+
+Implementation Details
+----------------------
 
 If the input dataframe is empty or too 'small' (where no modules are found),
 SPRAS will instead emit an empty output file.
diff --git a/docs/prms/meo.rst b/docs/prms/meo.rst
@@ -12,3 +12,16 @@ MEO takes in three optional parameters:
 * local_search: a "Yes"/"No" parameter that enables MEO's local search functionality. See "Improving approximations with local search" in
   the associated paper for more information. This should almost always be yes.
 * rand_restarts: the number (int) of random restarts to use.
+
+Dataset Usage
+-------------
+
+MEO uses ``sources``, ``targets``, and edge weights. MEO also uses
+edge direction information, and the output sub-network is directed.
+
+Implementation Details
+----------------------
+
+Internally, MEO only accepts directed interactomes.
+SPRAS will automatically convert edges to directed edges as necessary.
+For more information, see the section on :ref:`algorithm directionality <directionality>`.
diff --git a/docs/prms/mincostflow.rst b/docs/prms/mincostflow.rst
@@ -11,6 +11,19 @@ MinCostFlow takes two optional parameters:
 * flow: (int) the amount of flow going through the graph
 * capacity: the (float) max capacity for edges
 
+Dataset Usage
+-------------
+
+MinCostFlow uses the input's ``sources``, ``targets``, and edge weights. MinCostFlow also
+edge direction information.
+
+Implementation Details
+----------------------
+
+MinCostFlow's internal implementation only accepts directed interactomes.
+SPRAS will automatically convert edges to directed edges as necessary.
+For more information, see the section on :ref:`algorithm directionality <directionality>`.
+
 External links
 ++++++++++++++
 

diff --git a/docs/prms/oi1.rst b/docs/prms/oi1.rst
@@ -24,3 +24,12 @@ OI1 takes some optional arguments:
   * ``all``: connect the dummy node to all nodes in the interactome (i.e. full set of nodes in graph)
   * ``others``: connect the dummy node to all nodes that are not terminal nodes (i.e. nodes without prizes)
   * ``file``: connect the dummy node to a specific list of nodes provided in a file
+
+Dataset Usage
+-------------
+
+OmicsIntegrator1 prefers ``prize``s, but will take the union of ``sources`` and ``targets``
+and set their 'prize' to 1 if ``prize`` is not specified. If any ``dummy_nodes`` are specified,
+these are passed to OmicsIntegrator1 and can have their behavior configured with ``dummy_mode``.
+
+OmicsIntegrator1 uses edge direction information.
diff --git a/docs/prms/oi2.rst b/docs/prms/oi2.rst
@@ -18,3 +18,13 @@ OI2 takes a few optional arguments:
     * "others" = connect to all nodes except for terminals
     * "all" = connect to all nodes in the interactome.
 * seed: The random seed to use for this run.
+
+
+Dataset Usage
+-------------
+
+OmicsIntegrator2 prefers ``prize``s, but will take the union of ``sources`` and ``targets``
+and set their 'prize' to 1 if ``prize`` is not specified.
+
+OmicsIntegrator2 does not use edge direction information: all edges are treated as undirected,
+and the output sub-network is undirected.
diff --git a/docs/prms/pathlinker.rst b/docs/prms/pathlinker.rst
@@ -10,6 +10,20 @@ PathLinker takes one optional argument:
 
 * k: The number of paths to find (*k* shortest paths).
 
+Dataset Usage
+-------------
+
+PathLinker uses ``sources``, ``targets``, and edge weights. PathLinker
+uses edge direction information.
+
+Implementation Details
+----------------------
+
+Internally, PathLinker only takes in directed graphs.
+SPRAS will automatically convert edges to directed edges as necessary.
+For more information, see the section on :ref:`algorithm directionality <directionality>`.
+
+
 External links
 ++++++++++++++
 

diff --git a/docs/prms/prms.rst b/docs/prms/prms.rst
@@ -6,6 +6,29 @@ Pathway Reconstruction Methods
    there are other sub-types of algorithms that have more specialized capabilities than the rest.
    Currently, SPRAS also supports edge orientation algorithms (e.g. MEO) and active module identifiers/disease module mining methods (e.g. DOMINO).
 
+This is the list of SPRAS's supported pathway reconstruction methods. Each subpage comes with a description of the algorithm,
+its source code and associated paper (if one exists), and its 'dataset usage,' or parts of a dataset that it will utilize when
+running pathway reconstruction. Implementation details are also provided, for users wondering about any
+important decisions that differentiate the SPRAS-wrapped version from the original
+algorithm.
+
+.. _directionality:
+
+Directionality Details
+----------------------
+
+Some algorithms only accept fully undirected or fully directed interactomes as input. For input
+data, SPRAS will try to preserve as much directionality information as possible. Mixed interactomes
+are also accepted in SPRAS.
+
+SPRAS will automatically convert the input interactome to the desired directionality by the algorithm:
+this can mean that, for some algorithms, interactome direction may be ignored. Other algorithms will
+consider interactome directionality, whether by accepting mixed interactomes directly,
+or converting undirected edges into directed edges.
+
+For converting undirected edges to directed edges, unless otherwise specified, undirected edges
+will be converted into two directed edges pointing opposite of one another.
+
 .. toctree::
    :maxdepth: 1
    :caption: All Pairs

diff --git a/docs/prms/responsenet.rst b/docs/prms/responsenet.rst
@@ -9,3 +9,16 @@ https://github.com/Reed-CompBio/ResponseNet.
 ResponseNet takes one optional parameter:
 
 * gamma: (int) controls the size of the output graph: more gamma means more 'flow' gets passed along starting from the sources.
+
+Dataset Usage
+-------------
+
+ResponseNet uses ``sources``, ``targets``, and edge weights. ResponseNet
+uses edge direction information, but returns an undirected subnetwork.
+
+Implementation Details
+----------------------
+
+Internally, ResponseNet only takes in directed graphs.
+SPRAS will automatically convert edges to directed edges as necessary.
+For more information, see the section on :ref:`algorithm directionality <directionality>`.
diff --git a/docs/prms/rwr.rst b/docs/prms/rwr.rst
@@ -1,5 +1,5 @@
 RWR
-==========
+===
 
 RWR, or random walk with restarts, is a source and target independent pathway reconstruction algorithm
 that performs PathRank on the input interactome, using the provided prizes.
@@ -11,3 +11,20 @@ RWR takes in two parameters:
 * alpha: The damping factor of the internal PathRank algorithm. This is the probability that RWR randomly chooses a neighbor instead of restarting.
 
 RWR is implemented at https://github.com/reed-compbio/rwr.
+
+Dataset Usage
+-------------
+
+RWR considers the union of ``sources`` and ``targets`` as the
+input active nodes. Input interactome directionality is considered, and the
+output subnetwork is also directed.
+
+Implementation Details
+----------------------
+
+RWR returns a ranked list of nodes: SPRAS returns the induced subgraph
+from the number of nodes corresponding to the user-specified ``threshold``.
+
+Internally, RWR only takes in directed graphs.
+SPRAS will automatically convert edges to directed edges as necessary.
+For more information, see the section on :ref:`algorithm directionality <directionality>`.
diff --git a/docs/prms/strwr.rst b/docs/prms/strwr.rst
@@ -1,7 +1,7 @@
 ST_RWR
-==========
+======
 
-ST_RWR, or random walk with restarts, is a source and target dependent pathway reconstruction algorithm
+ST_RWR, or source-target random walk with restarts, is a source and target dependent pathway reconstruction algorithm
 that performs PathRank on the input interactome, using its edge weights, prizes, sources, and targets.
 
 For a random walk with restarts implementation that does not use  sources and targets, see RWR.
@@ -10,3 +10,19 @@ For a random walk with restarts implementation that does not use  sources and ta
 * alpha: The damping factor of the internal PathRank algorithm. This is the probability that RWR randomly chooses a neighbor instead of restarting.
 
 ST_RWR is implemented at https://github.com/reed-compbio/rwr.
+
+Dataset Usage
+-------------
+
+ST_RWR considers ``sources`` and ``targets``. ST_RWR considers interactome directionality,
+and the output subnetwork is also directed.
+
+Implementation Details
+----------------------
+
+ST_RWR returns a ranked list of nodes: SPRAS returns the induced subgraph
+from the number of nodes corresponding to the user-specified ``threshold``.
+
+Internally, ST_RWR only takes in directed graphs.
+SPRAS will automatically convert edges to directed edges as necessary.
+For more information, see the section on :ref:`algorithm directionality <directionality>`.