Merge pull request #551 from clara-parabricks/poa-gfa

[cudapoa] GFA output
NVIDIA-Genomics-Research · Dec 4, 2020 · 16d5272 · 16d5272
2 parents c68960e + 0e10f24
commit 16d5272
Show file tree

Hide file tree

Showing 4 changed files with 66 additions and 2 deletions.
diff --git a/common/base/include/claraparabricks/genomeworks/utils/graph.hpp b/common/base/include/claraparabricks/genomeworks/utils/graph.hpp
@@ -160,6 +160,19 @@ class Graph
         }
     }
 
+    ///
+    /// \brief Generates a GFA S(sequence) line from a node's label and sequence.
+    ///
+    /// \param gfa_str An output stringstream to write S lines to.
+    void node_labels_to_gfa(std::ostringstream& gfa_str) const
+    {
+        for (auto& iter : node_labels_)
+        {
+            gfa_str << "S"
+                    << "\t" << iter.first << "\t" << iter.second << std::endl;
+        }
+    }
+
     /// \brief Serialize edges to dot format
     ///
     /// \param dot_str Output string stream to serialize labels to
@@ -175,6 +188,27 @@ class Graph
         }
     }
 
+    ///
+    /// \brief Serialize edges to GFA (v1) format
+    ///
+    /// \param gfa_str An output stringstream to write output to.
+    void edges_to_gfa(std::ostringstream& gfa_str) const
+    {
+        for (auto& iter : edges_)
+        {
+            const edge_t& edge          = iter.first;
+            const node_id_t edge_source = edge.first;
+            const node_id_t edge_sink   = edge.second;
+            gfa_str << "L"
+                    << "\t" << edge_source << "\t"
+                    << "+"
+                    << "\t" << edge_sink << "\t"
+                    << "+"
+                    << "\t"
+                    << "*" << std::endl;
+        }
+    }
+
     /// List of adjacent nodes per node ID
     std::unordered_map<node_id_t, std::vector<node_id_t>> adjacent_nodes_;
 
@@ -214,6 +248,20 @@ class DirectedGraph : public Graph
         }
     }
 
+    /// \brief Serialize graph structure to GFA format
+    ///
+    /// \return A string encoding the graph in GFA format
+    std::string serialize_to_gfa() const
+    {
+        std::ostringstream gfa_str;
+        gfa_str << "H"
+                << "\t"
+                << "VN:Z:1.0" << std::endl;
+        node_labels_to_gfa(gfa_str);
+        edges_to_gfa(gfa_str);
+        return gfa_str.str();
+    }
+
     /// \brief Serialize graph structure to dot format
     ///
     /// \return A string encoding the graph in dot format

diff --git a/cudapoa/src/application_parameters.cpp b/cudapoa/src/application_parameters.cpp
@@ -43,6 +43,7 @@ ApplicationParameters::ApplicationParameters(int argc, char* argv[])
         {"graph-length", required_argument, 0, 'l'},
         {"pred-distance", required_argument, 0, 'D'},
         {"dot", required_argument, 0, 'd'},
+        {"gfa", required_argument, 0, 'G'},
         {"max-groups", required_argument, 0, 'M'},
         {"gpu-mem-alloc", required_argument, 0, 'R'},
         {"match", required_argument, 0, 'm'},
@@ -52,7 +53,7 @@ ApplicationParameters::ApplicationParameters(int argc, char* argv[])
         {"help", no_argument, 0, 'h'},
     };
 
-    std::string optstring = "i:ab:w:s:l:D:d:M:R:m:n:g:vh";
+    std::string optstring = "i:ab:w:s:l:D:d:G:M:R:m:n:g:vh";
 
     int32_t argument = 0;
     while ((argument = getopt_long(argc, argv, optstring.c_str(), options, nullptr)) != -1)
@@ -91,6 +92,10 @@ ApplicationParameters::ApplicationParameters(int argc, char* argv[])
         case 'd':
             graph_output_path = std::string(optarg);
             break;
+        case 'G':
+            graph_output_path = std::string(optarg);
+            output_gfa        = true;
+            break;
         case 'M':
             max_groups = std::stoi(optarg);
             break;
@@ -213,6 +218,9 @@ void ApplicationParameters::help(int32_t exit_code)
         -d, --dot <file>
             output path for printing graph in DOT format [disabled])"
               << R"(
+        -G, --gfa <file>
+            output path for printing graph in GFA format [disabled])"
+              << R"(
         -M, --max-groups  <int>
             maximum number of POA groups to create from file (-1 for all, > 0 for limited) [-1]
             repeats groups if less groups are present than specified)"

diff --git a/cudapoa/src/application_parameters.hpp b/cudapoa/src/application_parameters.hpp
@@ -39,6 +39,7 @@ class ApplicationParameters
 
     std::vector<std::string> input_paths;
     std::string graph_output_path;
+    bool output_gfa             = false;
     bool all_fasta              = true;
     bool msa                    = false; // consensus by default
     BandMode band_mode          = BandMode::adaptive_band;

diff --git a/cudapoa/src/main.cpp b/cudapoa/src/main.cpp
@@ -254,7 +254,14 @@ int main(int argc, char* argv[])
                         batch->get_graphs(graph, graph_status);
                         for (auto& g : graph)
                         {
-                            graph_output << g.serialize_to_dot() << std::endl;
+                            if (parameters.output_gfa)
+                            {
+                                graph_output << g.serialize_to_gfa() << std::endl;
+                            }
+                            else
+                            {
+                                graph_output << g.serialize_to_dot() << std::endl;
+                            }
                         }
                     }