From 00a1940377463072e30357bc76e4203c5d7ca8cd Mon Sep 17 00:00:00 2001
From: Marcin Wojdyr <wojdyr@gmail.com>
Date: Thu, 16 May 2024 20:47:51 +0200
Subject: [PATCH] to_mmcif: write _entity_poly_seq.hetero

writes y/n if the sequence was read from mmCIF _entity_poly_seq,
but unknown (?) if it's based on PDB SEQRES

I haven't tested the output on PDB deposition yet.
---
 include/gemmi/metadata.hpp |  2 ++
 src/mmcif.cpp              |  4 +++-
 src/to_mmcif.cpp           | 14 ++++++++------
 3 files changed, 13 insertions(+), 7 deletions(-)

diff --git a/include/gemmi/metadata.hpp b/include/gemmi/metadata.hpp
index 52b1873e..665b4f43 100644
--- a/include/gemmi/metadata.hpp
+++ b/include/gemmi/metadata.hpp
@@ -234,6 +234,8 @@ struct Entity {
   std::vector<std::string> subchains;
   EntityType entity_type = EntityType::Unknown;
   PolymerType polymer_type = PolymerType::Unknown;
+  // In case of microheterogeneity, PDB SEQRES has only the first residue name.
+  bool reflects_microhetero = false;
   std::vector<DbRef> dbrefs;
   /// List of SIFTS Uniprot ACs referenced by SiftsUnpResidue::acc_index
   std::vector<std::string> sifts_unp_acc;
diff --git a/src/mmcif.cpp b/src/mmcif.cpp
index 249560bd..e09c1040 100644
--- a/src/mmcif.cpp
+++ b/src/mmcif.cpp
@@ -814,6 +814,8 @@ Structure make_structure_from_block(const cif::Block& block_) {
         ent.polymer_type = polymer_type_from_string(poly_type);
       } catch (std::runtime_error&) {}
     }
+    // _entity_poly_seq is supposed to reflect heterogeneities in _atom_site.
+    ent.reflects_microhetero = true;
     st.entities.push_back(ent);
   }
 
@@ -825,7 +827,7 @@ Structure make_structure_from_block(const cif::Block& block_) {
       if (pos == (int) ent->full_sequence.size())
         ent->full_sequence.push_back(row.str(2));
       else if (pos >= 0 && pos < (int) ent->full_sequence.size())
-        ent->full_sequence[pos] += "," + row.str(2);
+        cat_to(ent->full_sequence[pos], ',', row.str(2));
     }
 
   cif::Table struct_ref = block.find("_struct_ref.",
diff --git a/src/to_mmcif.cpp b/src/to_mmcif.cpp
index 16936e6c..00d7825c 100644
--- a/src/to_mmcif.cpp
+++ b/src/to_mmcif.cpp
@@ -1106,23 +1106,25 @@ void update_mmcif_block(const Structure& st, cif::Block& block, MmcifOutputGroup
   }
 
   if (groups.entity_poly_seq) {
-    // SEQRES from PDB doesn't record microheterogeneity, so if the resulting
-    // cif has unknown("?") _entity_poly_seq.num, it cannot be trusted.
     cif::Loop& poly_loop = block.init_mmcif_loop("_entity_poly_seq.",
-                                           {"entity_id", "num", "mon_id"});
+                                     {"entity_id", "num", "mon_id", "hetero"});
     for (const Entity& ent : st.entities)
-      if (ent.entity_type == EntityType::Polymer)
+      if (ent.entity_type == EntityType::Polymer) {
+        // SEQRES from PDB doesn't record microheterogeneity.
+        std::string hetero_no = ent.reflects_microhetero ? "n" : "?";
         for (size_t i = 0; i != ent.full_sequence.size(); ++i) {
           const std::string& mon_ids = ent.full_sequence[i];
           std::string num = std::to_string(i+1);
           size_t start = 0, end;
           while ((end = mon_ids.find(',', start)) != std::string::npos) {
             poly_loop.add_row({qchain(ent.name), num,
-                               mon_ids.substr(start, end-start)});
+                               mon_ids.substr(start, end-start), "y"});
             start = end + 1;
           }
-          poly_loop.add_row({qchain(ent.name), num, mon_ids.substr(start)});
+          poly_loop.add_row({qchain(ent.name), num, mon_ids.substr(start),
+                             start == 0 ? hetero_no : "y"});
         }
+      }
   }
 
   if (groups.atoms)