seqtree 0.5.0__tar.gz → 0.6.0__tar.gz

{seqtree-0.5.0 → seqtree-0.6.0}/CHANGELOG.md

@@ -3,6 +3,33 @@
 All notable changes to `seqtree`. Dates are release dates; the project is pre-1.0, so a **minor**
 bump may carry breaking changes.
 
+## [0.6.0] — 2026-07-17
+
+### Fixed
+
+- **`structural` scored A–N contacts as if they did not interact, because the source table is
+  corrupted there.** The Miyazawa–Jernigan A–N contact energy was transcribed as `0.00` — the
+  generator's comment excused it as a pair the source left unlisted. It is not unlisted. In
+  `MJ_Keskin_potentials.csv` the lower triangle runs `A-A`, `R-A`, `R-R`, `[N-A]`, `N-R`, …, and
+  the `N-A` slot reads `V,1` (mirrored `1,V`), where `1` is a mangled residue symbol. The true
+  value is **0.15**. It is not a stray duplicate of `V-N` either — the V row separately lists
+  `N = 0.12`.
+
+  Substituting `0.00` for `0.15` understated A's interaction strength and overstated N's
+  (`q(A)` −0.04100 → −0.03350, `q(N)` +0.04350 → +0.05100). Since `structural` is rank-1 by
+  construction — every cell is a function of the 20 per-residue strengths — this moved the
+  strong→weak ordering that the matrix exists to encode, swapping N and P:
+  `FWCLYMIVHGAT`**`NP`**`RSQDEK` → `FWCLYMIVHGAT`**`PN`**`RSQDEK`.
+
+  Four cells of the 24×24 grid change: `A-W`/`W-A` 6 → 5, and `B-C`/`C-B` 4 → 3 (`B` is
+  `mean(N, D)`, so N's shift carries into it). **Scores from `structural()` change for sequences
+  containing A, W, or B**; all other matrices are untouched. N and P were near-tied, which is why
+  a real correction to the ordering moves so few cells.
+
+- **The same comment claimed the source is "near- but not perfectly symmetric".** It is perfectly
+  symmetric — 0 asymmetric directed pairs across all 400. The symmetrisation in
+  `structural_grid()` is a no-op guard, not a repair, and is now documented as such.
+
 ## [0.5.0] — 2026-07-16
 
 ### Added

{seqtree-0.5.0 → seqtree-0.6.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: seqtree
-Version: 0.5.0
+Version: 0.6.0
 Summary: Fast fuzzy search over biological sequences (C++ core, Python bindings)
 Keywords: sequence-search,fuzzy-matching,CDR3,immunology,bioinformatics,trie
 Author-Email: ISALGO laboratory <mikhail.shugay@gmail.com>

{seqtree-0.5.0 → seqtree-0.6.0}/SOURCES.md

@@ -32,12 +32,18 @@ possible; and out-of-frame junctions escape thymic selection, making them a samp
 | | |
 |---|---|
 | provenance | **derived** (published log-odds / interaction-strength matrices) |
-| origin | BLOSUM62 (NCBI); EMBOSS EPAM250, EPAM100; Miyazawa–Jernigan interaction strengths |
+| origin | BLOSUM62 (NCBI); EMBOSS EPAM250, EPAM100; Miyazawa–Jernigan interaction strengths, transcribed into `MJ_CONTACT` in `bench/gen_matrices.py` from the `MJ` rows of `tcren-ms/src/tcren/data/MJ_Keskin_potentials.csv` (Miyazawa & Jernigan, J Mol Biol 1996, as tabulated by Keskin et al., Protein Sci 1998) |
 | regenerate | `python bench/gen_matrices.py` |
 
 Stored as non-negative penalties via the Gram transform `pen[a][b] = s[a][a] + s[b][b] − 2·s[a][b]`,
 clamped at zero, so `penalty(a, a) == 0` and the score defines a ball.
 
+**Known corruption in the MJ source.** `MJ_Keskin_potentials.csv` has no `A,N` row: the lower-triangle
+slot where `N-A` belongs is mangled to `V,1` (mirrored `1,V`) for *both* the `MJ` and `Keskin`
+potentials. `MJ_CONTACT` carries the correct `A-N = 0.15` from that row, so the constant will not
+diff cleanly against the CSV at that pair. The `Keskin` rows are unused here. Fix not yet upstreamed
+to `tcren-ms`.
+
 ## Downloaded on demand
 
 ### `load_control(name)` → `~/.cache/seqtree/control_*.sqtree`

{seqtree-0.5.0 → seqtree-0.6.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "scikit_build_core.build"
 
 [project]
 name = "seqtree"
-version = "0.5.0"
+version = "0.6.0"
 description = "Fast fuzzy search over biological sequences (C++ core, Python bindings)"
 readme = "README.md"
 requires-python = ">=3.10"

{seqtree-0.5.0 → seqtree-0.6.0}/src/structural.inc

@@ -4,11 +4,11 @@
 static const int32_t kStructuralSize = 24;
 static const int32_t kStructural[24 * 24] = {
 //    A    R    N    D    C    Q    E    G    H    I    L    K    M    F    P    S    T    W    Y    V    B    Z    X    *
-     10,   9,   9,   7,   6,   8,   7,  10,   9,   6,   6,   5,   6,   5,   9,   8,   9,   6,   6,   7,   8,   7,   7,   0,
+     10,   9,   9,   7,   6,   8,   7,  10,   9,   6,   6,   5,   6,   5,   9,   8,   9,   5,   6,   7,   8,   7,   7,   0,
       9,  10,  10,   8,   4,   9,   8,   8,   8,   5,   4,   6,   5,   4,  10,  10,   9,   4,   5,   6,   9,   9,   7,   0,
       9,  10,  10,   8,   4,   9,   8,   8,   8,   5,   4,   6,   5,   4,  10,  10,   9,   4,   5,   6,   9,   9,   7,   0,
       7,   8,   8,  10,   3,   9,  10,   7,   6,   3,   3,   8,   3,   2,   8,   9,   8,   3,   3,   4,   9,   9,   6,   0,
-      6,   4,   4,   3,  10,   4,   2,   6,   7,   9,  10,   1,   9,   9,   4,   4,   5,  10,  10,   8,   4,   3,   6,   0,
+      6,   4,   4,   3,  10,   4,   2,   6,   7,   9,  10,   1,   9,   9,   4,   4,   5,  10,  10,   8,   3,   3,   6,   0,
       8,   9,   9,   9,   4,  10,   9,   8,   7,   4,   4,   7,   4,   3,   9,  10,   9,   3,   4,   5,   9,   9,   7,   0,
       7,   8,   8,  10,   2,   9,  10,   6,   6,   3,   2,   8,   3,   2,   8,   8,   7,   2,   3,   4,   9,   9,   6,   0,
      10,   8,   8,   7,   6,   8,   6,  10,   9,   7,   6,   5,   6,   5,   8,   8,   9,   6,   6,   8,   8,   7,   7,   0,
@@ -21,10 +21,10 @@ static const int32_t kStructural[24 * 24] = {
       9,  10,  10,   8,   4,   9,   8,   8,   8,   5,   4,   6,   5,   4,  10,  10,   9,   4,   5,   6,   9,   9,   7,   0,
       8,  10,  10,   9,   4,  10,   8,   8,   7,   5,   4,   6,   5,   4,  10,  10,   9,   4,   4,   6,   9,   9,   7,   0,
       9,   9,   9,   8,   5,   9,   7,   9,   8,   6,   5,   6,   5,   4,   9,   9,  10,   5,   5,   7,   9,   8,   7,   0,
-      6,   4,   4,   3,  10,   3,   2,   6,   6,   9,  10,   0,   9,  10,   4,   4,   5,  10,   9,   8,   3,   3,   6,   0,
+      5,   4,   4,   3,  10,   3,   2,   6,   6,   9,  10,   0,   9,  10,   4,   4,   5,  10,   9,   8,   3,   3,   6,   0,
       6,   5,   5,   3,  10,   4,   3,   6,   7,  10,  10,   1,  10,   9,   5,   4,   5,   9,  10,   9,   4,   3,   6,   0,
       7,   6,   6,   4,   8,   5,   4,   8,   8,   9,   8,   2,   9,   8,   6,   6,   7,   8,   9,  10,   5,   5,   7,   0,
-      8,   9,   9,   9,   4,   9,   9,   8,   7,   4,   4,   7,   4,   3,   9,   9,   9,   3,   4,   5,   9,   9,   7,   0,
+      8,   9,   9,   9,   3,   9,   9,   8,   7,   4,   4,   7,   4,   3,   9,   9,   9,   3,   4,   5,   9,   9,   7,   0,
       7,   9,   9,   9,   3,   9,   9,   7,   6,   4,   3,   8,   4,   2,   9,   9,   8,   3,   3,   5,   9,   9,   6,   0,
       7,   7,   7,   6,   6,   7,   6,   7,   7,   7,   6,   4,   7,   6,   7,   7,   7,   6,   6,   7,   7,   6,   7,   0,
       0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,  10,