incomplete: protein k-mer support

ratschlab · May 30, 2024 · add9fff · add9fff
1 parent 0120fae
commit add9fff
Show file tree

Hide file tree

Showing 3 changed files with 31 additions and 5 deletions.
diff --git a/metagraph/src/graph/representation/hash/dbg_sshash.cpp b/metagraph/src/graph/representation/hash/dbg_sshash.cpp
@@ -2,7 +2,6 @@
 
 #include "common/seq_tools/reverse_complement.hpp"
 #include "common/threads/threading.hpp"
-#include "kmer/kmer_extractor.hpp"
 
 
 namespace mtg {
@@ -130,8 +129,12 @@ void DBGSSHash::map_to_nodes_with_rc(std::string_view sequence,
     for (size_t i = k_ - 1; i < sequence.size() && !terminate(); ++i) {
         uint_kmer.drop_char();
         uint_kmer.kth_char_or(k_ - 1, kmer_t::char_to_uint(sequence[i]));
+#if _PROTEIN_GRAPH
+        callback(sshash_to_graph_index(dict_.lookup_uint(uint_kmer)), false);
+#else
         auto res = dict_.lookup_advanced_uint(uint_kmer, true);
         callback(sshash_to_graph_index(res.kmer_id), res.kmer_orientation);
+#endif
     }
 }
 
@@ -140,19 +143,27 @@ DBGSSHash::node_index DBGSSHash::traverse(node_index node, char next_char) const
     kmer_t new_kmer = sshash::util::string_to_uint_kmer<kmer_t>(string_kmer.c_str(), k_);
     new_kmer.drop_char();
     new_kmer.kth_char_or(k_ - 1, kmer_t::char_to_uint(next_char));
+#if _PROTEIN_GRAPH
+    return sshash_to_graph_index(dict_.lookup_uint(new_kmer));
+#else
     auto res = dict_.lookup_advanced_uint(new_kmer, mode_ == CANONICAL);
     node_index next = sshash_to_graph_index(res.kmer_id);
     return res.kmer_orientation ? reverse_complement(next) : next;
+#endif
 }
 
 DBGSSHash::node_index DBGSSHash::traverse_back(node_index node, char prev_char) const {
     std::string string_kmer = get_node_sequence(node);
     kmer_t new_kmer = sshash::util::string_to_uint_kmer<kmer_t>(string_kmer.c_str(), k_);
     new_kmer.append_char(kmer_t::char_to_uint(prev_char));
     new_kmer.take_chars(k_);
+#if _PROTEIN_GRAPH
+    return sshash_to_graph_index(dict_.lookup_uint(new_kmer));
+#else
     auto res = dict_.lookup_advanced_uint(new_kmer, mode_ == CANONICAL);
     node_index prev = sshash_to_graph_index(res.kmer_id);
     return res.kmer_orientation ? reverse_complement(prev) : prev;
+#endif
 }
 
 void DBGSSHash::adjacent_outgoing_nodes(node_index node,
@@ -254,18 +265,26 @@ DBGSSHash::node_index DBGSSHash::kmer_to_node(std::string_view kmer) const {
     if (!num_nodes())
         return npos;
 
+#if _PROTEIN_GRAPH
+    return dict_.lookup(kmer.data());
+#else
     auto res = dict_.lookup_advanced(kmer.data(), mode_ == CANONICAL);
     node_index node = sshash_to_graph_index(res.kmer_id);
     return res.kmer_orientation ? reverse_complement(node) : node;
+#endif
 }
 
 std::pair<DBGSSHash::node_index, bool>
 DBGSSHash::kmer_to_node_with_rc(std::string_view kmer) const {
     if (!num_nodes())
         return std::make_pair(npos, false);
 
+#if _PROTEIN_GRAPH
+    return dict_.lookup(kmer.data());
+#else
     auto res = dict_.lookup_advanced(kmer.data(), true);
     return std::make_pair(sshash_to_graph_index(res.kmer_id), res.kmer_orientation);
+#endif
 }
 
 std::string DBGSSHash::get_node_sequence(node_index node) const {

diff --git a/metagraph/src/graph/representation/hash/dbg_sshash.hpp b/metagraph/src/graph/representation/hash/dbg_sshash.hpp
@@ -9,11 +9,18 @@
 #include "common/utils/string_utils.hpp"
 #include "common/logger.hpp"
 #include "graph/representation/base/sequence_graph.hpp"
+#include "kmer/kmer_extractor.hpp"
 
 namespace mtg::graph {
 
 class DBGSSHash : public DeBruijnGraph {
+#if _PROTEIN_GRAPH
+    using kmer_t = sshash::alpha_kmer_t<uint64_t,
+                                        kmer::KmerExtractor2Bit::bits_per_char,
+                                        kmer::alphabets::kAlphabetProtein>;
+#else
     using kmer_t = sshash::dna_uint_kmer_t<uint64_t>;
+#endif
 
   public:
     explicit DBGSSHash(size_t k, Mode mode = BASIC);

diff --git a/metagraph/src/kmer/alphabets.hpp b/metagraph/src/kmer/alphabets.hpp
@@ -87,7 +87,7 @@ static_assert(kBOSSSigmaDNA5 > 1llu << (kBOSSBitsPerCharDNA5 - 1));
 
 
 
-constexpr char kAlphabetProtein[] = "ABCDEFGHIJKLMNOPQRSTUVWYZX";
+inline constexpr char kAlphabetProtein[] = "ABCDEFGHIJKLMNOPQRSTUVWYZX";
 constexpr uint8_t kSigmaProtein = sizeof(kAlphabetProtein) - 1;
 constexpr uint8_t kBitsPerCharProtein = log2<kSigmaProtein - 1>::value + 1;
 constexpr uint8_t kCharToProtein[128] = {
@@ -106,7 +106,7 @@ static_assert(kSigmaProtein > 1llu << (kBitsPerCharProtein - 1));
 
 
 //for case-specific DNA and RNA (U <-> T) data
-constexpr char kAlphabetDNACaseSent[] = "ACGTNacgt";
+inline constexpr char kAlphabetDNACaseSent[] = "ACGTNacgt";
 constexpr uint8_t kSigmaDNACaseSent = sizeof(kAlphabetDNACaseSent) - 1;
 constexpr uint8_t kBitsPerCharDNACaseSent = log2<kSigmaDNACaseSent - 1>::value + 1;
 constexpr uint8_t kCharToDNACaseSent[128] = {
@@ -125,7 +125,7 @@ static_assert(kSigmaDNACaseSent > 1llu << (kBitsPerCharDNACaseSent - 1));
 
 
 //for DNA and RNA (U <-> T) alphabets
-constexpr char kAlphabetDNA[] = "ACGT";
+inline constexpr char kAlphabetDNA[] = "ACGT";
 constexpr uint8_t kSigmaDNA = sizeof(kAlphabetDNA) - 1;
 constexpr uint8_t kBitsPerCharDNA = log2<kSigmaDNA - 1>::value + 1;
 constexpr uint8_t kCharToDNA[128] = {
@@ -142,7 +142,7 @@ const std::vector<uint8_t> kComplementMapDNA = { 3, 2, 1, 0, 4 };
 static_assert(kSigmaDNA <= 1llu << kBitsPerCharDNA);
 static_assert(kSigmaDNA > 1llu << (kBitsPerCharDNA - 1));
 
-constexpr char kAlphabetDNA5[] = "ACGTN";
+inline constexpr char kAlphabetDNA5[] = "ACGTN";
 constexpr uint8_t kSigmaDNA5 = sizeof(kAlphabetDNA5) - 1;
 constexpr uint8_t kBitsPerCharDNA5 = log2<kSigmaDNA5 - 1>::value + 1;
 static_assert(kSigmaDNA5 <= 1llu << kBitsPerCharDNA5);