| Title: | An interface to the Semantic MEDLINE database |
|---|---|
| Description: | A programmatic interface to the Semantic MEDLINE database. It provides functions for searching the database for concepts and finding paths between concepts. Path searching can also be tailored to user specifications, such as placing restrictions on concept types and the type of link between concepts. It also provides functions for summarizing and visualizing those paths. |
| Authors: | Leslie Myint [aut, cre] |
| Maintainer: | Leslie Myint <[email protected]> |
| License: | Artistic-2.0 |
| Version: | 1.17.0 |
| Built: | 2024-11-26 06:08:40 UTC |
| Source: | https://github.com/bioc/rsemmed |
Search for nodes by name (exact match or using regular expressions)
or which match supplied semantic types. Perform anti-matching by
setting match = FALSE. Capitalization is ignored.
find_nodes(obj, pattern = NULL, names = NULL, semtypes = NULL, match = TRUE)find_nodes(obj, pattern = NULL, names = NULL, semtypes = NULL, match = TRUE)
obj |
Either the SemMed graph or a node set ( |
pattern |
Regular expression used to find matches in node names |
names |
Character vector of exact node names |
semtypes |
Character vector of semantic types |
match |
If |
A vertex sequence of matching nodes
data(g_mini) find_nodes(g_mini, pattern = "cortisol") find_nodes(g_mini, pattern = "cortisol$") find_nodes(g_mini, pattern = "stress") find_nodes(g_mini, pattern = "stress") %>% find_nodes(pattern = "disorder", match = FALSE) find_nodes(g_mini, names = "Serum cortisol") find_nodes(g_mini, names = "Chronic Stress") find_nodes(g_mini, semtypes = "dsyn") find_nodes(g_mini, semtypes = c("dsyn", "fndg")) ## pattern and semtypes are combined via OR: find_nodes(g_mini, pattern = "cortisol", semtypes = "horm") ## To make an AND query, chain find_nodes sequenctially: find_nodes(g_mini, pattern = "cortisol") %>% find_nodes(semtypes = "horm")data(g_mini) find_nodes(g_mini, pattern = "cortisol") find_nodes(g_mini, pattern = "cortisol$") find_nodes(g_mini, pattern = "stress") find_nodes(g_mini, pattern = "stress") %>% find_nodes(pattern = "disorder", match = FALSE) find_nodes(g_mini, names = "Serum cortisol") find_nodes(g_mini, names = "Chronic Stress") find_nodes(g_mini, semtypes = "dsyn") find_nodes(g_mini, semtypes = c("dsyn", "fndg")) ## pattern and semtypes are combined via OR: find_nodes(g_mini, pattern = "cortisol", semtypes = "horm") ## To make an AND query, chain find_nodes sequenctially: find_nodes(g_mini, pattern = "cortisol") %>% find_nodes(semtypes = "horm")
Find all shortest paths between sets of nodes
find_paths(graph, from, to, weights = NULL)find_paths(graph, from, to, weights = NULL)
graph |
The SemMed graph |
from |
A set of source nodes. |
to |
A set of destination nodes. |
weights |
A numeric vector of edge weights. If |
find_paths relies on igraph::all_shortest_paths to find all
shortest paths between the nodes in from and to. This
function searches for undirected paths.
Because the Semantic MEDLINE graph is a multigraph, there may be multiple
paths with the same sequence of nodes. This function collapses these into
a single node sequence. The display functions (text_path and
plot_path) take care of showing the multiple edges leading to
repeated paths.
A list of shortest paths. List items correspond to the
node(s) given in from.
make_edge_weights to tailor the
shortest path search
data(g_mini) node_cortisol <- find_nodes(g_mini, names = "Serum cortisol") node_stress <- find_nodes(g_mini, names = "Chronic Stress") find_paths(g_mini, from = node_cortisol, to = node_stress)data(g_mini) node_cortisol <- find_nodes(g_mini, names = "Serum cortisol") node_stress <- find_nodes(g_mini, names = "Chronic Stress") find_paths(g_mini, from = node_cortisol, to = node_stress)
A dataset containing a very small subset of the full Semantic MEDLINE graph.
data(g_mini)data(g_mini)
An igraph with 7 nodes and 15 edges
A dataset containing a small subset of the full Semantic MEDLINE graph.
data(g_small)data(g_small)
An igraph with 1038 nodes and 318,105 edges
Search for nodes by name using regular expressions or which match given
semantic types. Perform anti-matching by setting match = FALSE.
get_edge_features( graph, include_degree = FALSE, include_node_ids = FALSE, include_num_instances = FALSE )get_edge_features( graph, include_degree = FALSE, include_node_ids = FALSE, include_num_instances = FALSE )
graph |
The SemMed graph |
include_degree |
If |
include_node_ids |
If |
include_num_instances |
If |
A tbl where each row corresponds to an edge in the
Semantic MEDLINE graph. The ordering of the rows corresponds to
E(graph). Features (columns) always returned include the
name and semantic type of the head (subject) and tail (object) nodes.
make_edge_weights for using this data to
construct edge weights
data(g_mini) e_feat <- get_edge_features(g_mini)data(g_mini) e_feat <- get_edge_features(g_mini)
For each pair of source and target nodes in object, obtain the
names of middle nodes on paths.
get_middle_nodes(graph, object, collapse = TRUE)get_middle_nodes(graph, object, collapse = TRUE)
graph |
The SemMed graph |
object |
A vertex sequence ( |
collapse |
If |
A tbl where each row corresponds to a source-target pair
in object. The last column is a list-column containing character
vectors of names of middle nodes.
data(g_mini) node_cortisol <- find_nodes(g_mini, "Serum cortisol") node_stress <- find_nodes(g_mini, "Chronic Stress") paths <- find_paths(g_mini, from = node_cortisol, to = node_stress) middle <- get_middle_nodes(g_mini, paths)data(g_mini) node_cortisol <- find_nodes(g_mini, "Serum cortisol") node_stress <- find_nodes(g_mini, "Chronic Stress") paths <- find_paths(g_mini, from = node_cortisol, to = node_stress) middle <- get_middle_nodes(g_mini, paths)
Grow a set of nodes into its first order neighborhood.
grow_nodes(graph, nodes)grow_nodes(graph, nodes)
graph |
The SemMed graph |
nodes |
A vertex sequence ( |
grow_nodes obtains the set of immediate neighbors of the
supplied nodes using igraph::ego. Unlike ego,
grow_nodes flattens the result from a list to an ordinary
vertex sequence and removes the original search nodes.
A vertex sequence of nodes in the neighborhood (not including the original nodes)
find_nodes for filtering out irrelevant
nodes from this set.
data(g_mini) node_cortisol <- find_nodes(g_mini, name = "hypercortisolemia") nbrs <- grow_nodes(g_mini, node_cortisol)data(g_mini) node_cortisol <- find_nodes(g_mini, name = "hypercortisolemia") nbrs <- grow_nodes(g_mini, node_cortisol)
Create edge weights to modify the shortest path search
(find_paths). Discourage and/or encourage certain types of paths
by supplying _out and _in arguments, respectively. Node
semantic types, node names, and edge predicates are the features that
can influence the edge weights. Capitalization is ignored.
make_edge_weights( graph, e_feat, node_semtypes_out = NULL, node_names_out = NULL, edge_preds_out = NULL, node_semtypes_in = NULL, node_names_in = NULL, edge_preds_in = NULL )make_edge_weights( graph, e_feat, node_semtypes_out = NULL, node_names_out = NULL, edge_preds_out = NULL, node_semtypes_in = NULL, node_names_in = NULL, edge_preds_in = NULL )
graph |
The SemMed graph |
e_feat |
A |
node_semtypes_out |
A character vector of semantic types to exclude from shortest paths. |
node_names_out |
A character vector of exact node names to exclude. |
edge_preds_out |
A character vector of edge predicates to exclude. |
node_semtypes_in |
A character vector of semantic types to include/encourage in shortest paths. |
node_names_in |
A character vector of exact node names to include. |
edge_preds_in |
A character vector of edge predicates to include. |
A numeric vector of weights
find_paths, get_middle_nodes for a
way to obtain node names to remove
data(g_mini) node_cortisol <- find_nodes(g_mini, names = "Serum cortisol") node_stress <- find_nodes(g_mini, names = "Chronic Stress") paths <- find_paths(g_mini, from = node_cortisol, to = node_stress) e_feat <- get_edge_features(g_mini) w1 <- make_edge_weights(g_mini, e_feat, edge_preds_in = "COEXISTS_WITH") paths1 <- find_paths(g_mini, from = node_cortisol, to = node_stress, weights = w1) w2 <- make_edge_weights(g_mini, e_feat, edge_preds_in = "ISA", node_names_out = "Stress") paths2 <- find_paths(g_mini, from = node_cortisol, to = node_stress, weights = w2)data(g_mini) node_cortisol <- find_nodes(g_mini, names = "Serum cortisol") node_stress <- find_nodes(g_mini, names = "Chronic Stress") paths <- find_paths(g_mini, from = node_cortisol, to = node_stress) e_feat <- get_edge_features(g_mini) w1 <- make_edge_weights(g_mini, e_feat, edge_preds_in = "COEXISTS_WITH") paths1 <- find_paths(g_mini, from = node_cortisol, to = node_stress, weights = w1) w2 <- make_edge_weights(g_mini, e_feat, edge_preds_in = "ISA", node_names_out = "Stress") paths2 <- find_paths(g_mini, from = node_cortisol, to = node_stress, weights = w2)
Plot the graph form of a path
plot_path(graph, path)plot_path(graph, path)
graph |
The SemMed graph |
path |
A vertex sequence ( |
All connections among nodes along the supplied path are plotted with nodes labeled with their name and edges labeled with their predicate.
A plot is created on the current graphics device
text_path for textual display of paths
data(g_mini) node_cortisol <- find_nodes(g_mini, names = "Serum cortisol") node_stress <- find_nodes(g_mini, names = "Chronic Stress") paths <- find_paths(g_mini, from = node_cortisol, to = node_stress) plot_path(g_mini, paths[[1]][[1]])data(g_mini) node_cortisol <- find_nodes(g_mini, names = "Serum cortisol") node_stress <- find_nodes(g_mini, names = "Chronic Stress") paths <- find_paths(g_mini, from = node_cortisol, to = node_stress) plot_path(g_mini, paths[[1]][[1]])
Summarize the predicates present in a collection of paths
summarize_predicates(graph, object, print = TRUE)summarize_predicates(graph, object, print = TRUE)
graph |
The SemMed graph |
object |
A vertex sequence ( |
print |
If |
Because predicates are edge features, it is assumed that by using
summarize_predicates the nodes contained in object are
ordered (paths). This is why summarize_semtypes has the
is_path argument, but summarize_predicates does not.
summarize_predicates tabulates edge predicates across paths
corresponding to each from-to pair in object.
A tbl where each row corresponds to a
from-to pair in object. The last column is a
list-column containing table's of predicate counts.
summarize_semtypes for tabulating
semantic types of nodes in paths or other node collections
data(g_mini) node_cortisol <- find_nodes(g_mini, "Serum cortisol") node_stress <- find_nodes(g_mini, "Chronic Stress") paths <- find_paths(g_mini, from = node_cortisol, to = node_stress) summarize_predicates(g_mini, paths)data(g_mini) node_cortisol <- find_nodes(g_mini, "Serum cortisol") node_stress <- find_nodes(g_mini, "Chronic Stress") paths <- find_paths(g_mini, from = node_cortisol, to = node_stress) summarize_predicates(g_mini, paths)
Summarize the semantic types present in a collection of nodes
summarize_semtypes(graph, object, print = TRUE, is_path = TRUE)summarize_semtypes(graph, object, print = TRUE, is_path = TRUE)
graph |
The SemMed graph |
object |
A vertex sequence ( |
print |
If |
is_path |
If |
summarize_semtypes summarizes the semantic types present in
supplied node collections and has different behavior depending on
whether the node collection is ordered (paths) or unordered. Using
is_path = TRUE indicates that the nodes are ordered. Using
is_path = FALSE indicates that the nodes are an unordered
collection, often from find_nodes or grow_nodes.
Using is_path = TRUE: When the node collection is ordered, the
object is assumed to be the result of find_paths or a subset of
such an object. Because find_paths returns a list of paths lists,
summarize_semtypes takes a single path, a list of paths, or a
list of path lists as input. In the case of a collection of ordered nodes,
summarize_semtypes counts the semantic types present in
object. If a node is associated with multiple semantic types,
each type is counted once. The first and last nodes of each path are
removed they correspond to the nodes in from and to from
find_paths, and it is assumed that the middle nodes on the paths
are more of interest. The tabulations are printed to screen (if
print = TRUE) and returned as table's. These table's
are bundled into a list-column of a tbl in the (invisbly returned)
output. Each row of the tbl corresponds to a from-to
pair present in object.
Using is_path = FALSE: This option is for summarizing results from
find_nodes and grow_nodes, which return unordered node sets.
(Note: paths and unordered node sets are both represented as igraph
vertex sequences (class igraph.vs).) The printed output shows
information for each semantic type present in object. It shows all
nodes of that semantic type as well as their degree and degree percentile
within the entire graph. The (invisibly returned) output combines
all of the printed information in a tbl.
Output is returned invisibly.
If is_path = TRUE, a tbl where each row corresponds
to a from-to pair in object. The last column
is a list-column containing table's of semantic type counts.
If is_path = FALSE, a tbl where each row corresponds
to a name-semantic type combination. Columns give node name,
semantic type, degree, and degree percentile.
summarize_predicates for summarizing
predicates on edges
find_paths for searching for paths
between node sets
find_nodes and grow_nodes
for searching for and filtering nodes
data(g_mini) node_cortisol <- find_nodes(g_mini, "Serum cortisol") node_stress <- find_nodes(g_mini, "Chronic Stress") paths <- find_paths(g_mini, from = node_cortisol, to = node_stress) summarize_semtypes(g_mini, paths) nodes_mood <- find_nodes(g_mini, "mood") summarize_semtypes(g_mini, nodes_mood, is_path = FALSE)data(g_mini) node_cortisol <- find_nodes(g_mini, "Serum cortisol") node_stress <- find_nodes(g_mini, "Chronic Stress") paths <- find_paths(g_mini, from = node_cortisol, to = node_stress) summarize_semtypes(g_mini, paths) nodes_mood <- find_nodes(g_mini, "mood") summarize_semtypes(g_mini, nodes_mood, is_path = FALSE)
Show a text display of a path and obtain output that can be used to explore predications along the path. (A predication is a SUBJECT–LINKING VERB–>OBJECT triple.)
text_path(graph, path, print = TRUE)text_path(graph, path, print = TRUE)
graph |
The SemMed graph |
path |
A vertex sequence ( |
print |
Print the path to screen? |
text_path invisibly returns a list of tbl's containing
information on the predications on the path. Each list element is a
tbl that corresponds to a (sequential) pair of nodes along
the path. The tbl contains information on the subject and
object node's name and semantic type as well as all predicates linking
the subject and object.
Invisibly returns a list of predications for each pair of nodes along the path.
plot_path for plotting paths
data(g_mini) node_cortisol <- find_nodes(g_mini, names = "Serum cortisol") node_stress <- find_nodes(g_mini, names = "Chronic Stress") paths <- find_paths(g_mini, from = node_cortisol, to = node_stress) text_path(g_mini, paths[[1]][[1]]) result <- text_path(g_mini, paths[[1]][[1]], print = FALSE)data(g_mini) node_cortisol <- find_nodes(g_mini, names = "Serum cortisol") node_stress <- find_nodes(g_mini, names = "Chronic Stress") paths <- find_paths(g_mini, from = node_cortisol, to = node_stress) text_path(g_mini, paths[[1]][[1]]) result <- text_path(g_mini, paths[[1]][[1]], print = FALSE)