sci-form 0.15.2

High-performance 3D molecular conformer generation using ETKDG distance geometry
Documentation
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
#![allow(
    unused_imports,
    unused_variables,
    dead_code,
    clippy::unnecessary_cast,
    clippy::needless_range_loop,
    clippy::manual_repeat_n,
    clippy::manual_str_repeat,
    clippy::manual_is_multiple_of,
    clippy::redundant_field_names,
    clippy::useless_vec,
    clippy::single_range_in_vec_init
)]
//! Step-by-step pipeline comparison for a failing molecule.
//! Loads RDKit's bounds matrix and final coordinates, then runs our pipeline
//! and prints diagnostics at each step to find where divergence occurs.

use nalgebra::DMatrix;
use sci_form::distgeom::{
    calculate_bounds_matrix_opts, compute_initial_coords_rdkit, identify_chiral_sets,
    identify_tetrahedral_centers, pick_rdkit_distances, triangle_smooth_tol, MinstdRand,
};
use sci_form::forcefield::bounds_ff::minimize_bfgs_rdkit;
use sci_form::forcefield::etkdg_3d::{build_etkdg_3d_ff_with_torsions, minimize_etkdg_3d_bfgs};

fn load_npy_f64(path: &str) -> DMatrix<f64> {
    let data = std::fs::read(path).expect("Failed to read npy file");
    // Parse numpy .npy format: 10-byte magic, then header, then data
    // Find the end of header (terminated by \n)
    let header_end = data.iter().position(|&b| b == b'\n').unwrap();
    // Parse shape from header
    let header = std::str::from_utf8(&data[10..header_end]).unwrap();
    let shape_start = header.find("(").unwrap() + 1;
    let shape_end = header.find(")").unwrap();
    let shape_str = &header[shape_start..shape_end];
    let dims: Vec<usize> = shape_str
        .split(',')
        .filter(|s| !s.trim().is_empty())
        .map(|s| s.trim().parse().unwrap())
        .collect();
    let nrows = dims[0];
    let ncols = if dims.len() > 1 { dims[1] } else { 1 };

    let data_start = header_end + 1;
    let f64_data: Vec<f64> = data[data_start..]
        .chunks(8)
        .map(|chunk| f64::from_le_bytes(chunk.try_into().unwrap()))
        .collect();

    // numpy stores row-major, DMatrix is column-major
    let mut mat = DMatrix::zeros(nrows, ncols);
    for i in 0..nrows {
        for j in 0..ncols {
            mat[(i, j)] = f64_data[i * ncols + j];
        }
    }
    mat
}

#[test]
fn test_trace_failing_molecule() {
    let smiles = "C#CCOC(C)CC1CC2C3CCC(C)C(O)(C3)C2O1";

    // Build molecule from reference JSON (matching RDKit's graph exactly)
    let ref_data =
        sci_form::fixture_io::read_text_fixture("tests/fixtures/gdb20_reference_1k.json")
            .expect("Reference file needed");
    let ref_mols: Vec<serde_json::Value> = serde_json::from_str(&ref_data).unwrap();
    let ref_mol = ref_mols
        .iter()
        .find(|m| m["smiles"].as_str().unwrap() == smiles)
        .expect("Molecule not found in reference");

    let mut mol = sci_form::graph::Molecule::new(smiles);
    let atoms = ref_mol["atoms"].as_array().unwrap();
    let bonds = ref_mol["bonds"].as_array().unwrap();
    let mut node_indices = Vec::new();
    for atom in atoms {
        let element = atom["element"].as_u64().unwrap() as u8;
        let hyb_str = atom["hybridization"].as_str().unwrap();
        let hybridization = match hyb_str {
            "SP" => sci_form::graph::Hybridization::SP,
            "SP2" => sci_form::graph::Hybridization::SP2,
            "SP3" => sci_form::graph::Hybridization::SP3,
            "SP3D" => sci_form::graph::Hybridization::SP3D,
            "SP3D2" => sci_form::graph::Hybridization::SP3D2,
            _ => sci_form::graph::Hybridization::Unknown,
        };
        let fc = atom["formal_charge"].as_i64().unwrap_or(0) as i8;
        let new_atom = sci_form::graph::Atom {
            element,
            position: nalgebra::Vector3::zeros(),
            charge: 0.0,
            formal_charge: fc,
            hybridization,
            chiral_tag: sci_form::graph::ChiralType::Unspecified,
            explicit_h: if element == 1 || element == 0 { 1 } else { 0 },
        };
        node_indices.push(mol.add_atom(new_atom));
    }
    for bond in bonds {
        let start = bond["start"].as_u64().unwrap() as usize;
        let end = bond["end"].as_u64().unwrap() as usize;
        let order_str = bond["order"].as_str().unwrap();
        let order = match order_str {
            "DOUBLE" => sci_form::graph::BondOrder::Double,
            "TRIPLE" => sci_form::graph::BondOrder::Triple,
            "AROMATIC" => sci_form::graph::BondOrder::Aromatic,
            _ => sci_form::graph::BondOrder::Single,
        };
        mol.add_bond(
            node_indices[start],
            node_indices[end],
            sci_form::graph::Bond {
                order,
                stereo: sci_form::graph::BondStereo::None,
            },
        );
    }

    let n = mol.graph.node_count();
    println!("\nMolecule: {} (built from reference JSON)", smiles);
    println!("Atoms: {}", n);

    // Load RDKit bounds
    let rdkit_bounds = load_npy_f64("/tmp/rdkit_bounds_fail.npy");
    println!(
        "RDKit bounds loaded: {}x{}",
        rdkit_bounds.nrows(),
        rdkit_bounds.ncols()
    );

    // Our bounds
    let raw = calculate_bounds_matrix_opts(&mol, true);
    let mut bounds = raw;
    let ok = triangle_smooth_tol(&mut bounds, 0.0);
    println!("Our smoothing OK: {}", ok);

    // Compare bounds
    let mut max_ub_diff = 0.0f64;
    let mut max_lb_diff = 0.0f64;
    let mut ub_diff_count = 0;
    let mut lb_diff_count = 0;
    for i in 0..n {
        for j in (i + 1)..n {
            let our_ub = bounds[(i, j)];
            let rdkit_ub = rdkit_bounds[(i, j)];
            let diff = (our_ub - rdkit_ub).abs();
            if diff > 1e-10 {
                ub_diff_count += 1;
                if diff > max_ub_diff {
                    max_ub_diff = diff;
                }
            }
            let our_lb = bounds[(j, i)];
            let rdkit_lb = rdkit_bounds[(j, i)];
            let diff = (our_lb - rdkit_lb).abs();
            if diff > 1e-10 {
                lb_diff_count += 1;
                if diff > max_lb_diff {
                    max_lb_diff = diff;
                }
            }
        }
    }
    println!("\n=== Bounds Comparison ===");
    println!(
        "UB diffs (>1e-10): {} max: {:.2e}",
        ub_diff_count, max_ub_diff
    );
    println!(
        "LB diffs (>1e-10): {} max: {:.2e}",
        lb_diff_count, max_lb_diff
    );

    // Run our pipeline
    let chiral_sets = identify_chiral_sets(&mol);
    let _tet_centers = identify_tetrahedral_centers(&mol);
    let use_4d = !chiral_sets.is_empty();
    let embed_dim = if use_4d { 4 } else { 3 };
    println!(
        "\nChiral sets: {}, use_4d: {}, embed_dim: {}",
        chiral_sets.len(),
        use_4d,
        embed_dim
    );

    let mut rng = MinstdRand::new(42);

    // Step 1: Random distances
    let dists = pick_rdkit_distances(&mut rng, &bounds);
    println!("\n=== Random Distances (first 10 pairs) ===");
    for i in 0..n.min(10) {
        for j in (i + 1)..n.min(i + 3) {
            println!("  d({},{}) = {:.10}", i, j, dists[(i, j)]);
        }
    }

    // Also compute what distances RDKit would generate using same bounds
    let mut rng2 = MinstdRand::new(42);
    let dists_from_rdkit_bounds = pick_rdkit_distances(&mut rng2, &rdkit_bounds);
    println!("\n=== Distance comparison (our bounds vs RDKit bounds, first mismatches) ===");
    let mut dist_diff_count = 0;
    let mut max_dist_diff = 0.0f64;
    for i in 0..n {
        for j in (i + 1)..n {
            let diff = (dists[(i, j)] - dists_from_rdkit_bounds[(i, j)]).abs();
            if diff > 1e-15 {
                dist_diff_count += 1;
                if diff > max_dist_diff {
                    max_dist_diff = diff;
                }
                if dist_diff_count <= 5 {
                    println!(
                        "  d({},{}) ours={:.12} rdkit_b={:.12} diff={:.2e}",
                        i,
                        j,
                        dists[(i, j)],
                        dists_from_rdkit_bounds[(i, j)],
                        diff
                    );
                }
            }
        }
    }
    println!(
        "Total distance diffs: {} max: {:.2e}",
        dist_diff_count, max_dist_diff
    );

    // Step 2: Compute metric matrix and eigen decomposition (to print eigenvalues)
    {
        let n = mol.graph.node_count();
        let d_size = n * (n + 1) / 2;
        let mut sq_packed = vec![0.0f64; d_size];
        let mut sum_sq_all = 0.0f64;
        for i in 0..n {
            let id = i * (i + 1) / 2;
            for j in 0..=i {
                let d = dists[(i, j)];
                sq_packed[id + j] = d * d;
                sum_sq_all += d * d;
            }
        }
        sum_sq_all /= (n * n) as f64;

        let mut d0 = vec![0.0f64; n];
        for i in 0..n {
            let mut row_sum = 0.0f64;
            for j in 0..n {
                let idx = if i >= j {
                    i * (i + 1) / 2 + j
                } else {
                    j * (j + 1) / 2 + i
                };
                row_sum += sq_packed[idx];
            }
            d0[i] = row_sum / n as f64 - sum_sq_all;
        }

        let mut t_packed = vec![0.0f64; d_size];
        for i in 0..n {
            let id = i * (i + 1) / 2;
            for j in 0..=i {
                let sq_val = sq_packed[if i >= j {
                    i * (i + 1) / 2 + j
                } else {
                    j * (j + 1) / 2 + i
                }];
                t_packed[id + j] = 0.5 * (d0[i] + d0[j] - sq_val);
            }
        }

        let eigen_seed = (sum_sq_all * n as f64) as i32;
        println!("\n=== Metric Matrix Diagnostics ===");
        println!("sum_sq_all = {:.10}", sum_sq_all);
        println!("eigen_seed = {}", eigen_seed);
        println!(
            "d0[0]={:.10}, d0[1]={:.10}, d0[2]={:.10}",
            d0[0], d0[1], d0[2]
        );
        println!(
            "t_packed[46]={:.10} t_packed[47]={:.10}",
            t_packed[46], t_packed[47]
        );

        let result = sci_form::distgeom::power_eigen_solver(3, &mut t_packed, n, eigen_seed);
        match result {
            Some((eigenvalues, eigenvectors)) => {
                for (i, ev) in eigenvalues.iter().enumerate() {
                    println!(
                        "  λ{} = {:.10}  v[0..3] = [{:.10}, {:.10}, {:.10}]",
                        i, ev, eigenvectors[i][0], eigenvectors[i][1], eigenvectors[i][2]
                    );
                }
            }
            None => println!("  Power iteration FAILED!"),
        }
    }

    // Step 2b: Initial coordinates
    let coords_opt = compute_initial_coords_rdkit(&mut rng, &dists, embed_dim);
    let mut coords = match coords_opt {
        Some(c) => c,
        None => {
            println!("EMBEDDING FAILED!");
            return;
        }
    };
    println!("\n=== Initial Coordinates (first 5 atoms) ===");
    for i in 0..n.min(5) {
        println!(
            "  {}: ({:.8}, {:.8}, {:.8})",
            i,
            coords[(i, 0)],
            coords[(i, 1)],
            coords[(i, 2)]
        );
    }

    // Step 3: First minimization (bounds FF)
    let basin = 5.0f32;
    let force_tol = 1e-3f64;
    let mut need_more = 1;
    let mut iters = 0;
    while need_more != 0 {
        need_more = minimize_bfgs_rdkit(
            &mut coords,
            &bounds,
            &chiral_sets,
            400,
            force_tol,
            basin,
            0.1,
            1.0,
        );
        iters += 1;
    }
    println!("\n=== After Bounds FF (stage 1, {} rounds) ===", iters);
    for i in 0..n.min(5) {
        println!(
            "  {}: ({:.8}, {:.8}, {:.8})",
            i,
            coords[(i, 0)],
            coords[(i, 1)],
            coords[(i, 2)]
        );
    }

    // Step 4: Check energy
    let e_total = sci_form::forcefield::bounds_ff::bounds_violation_energy_basin(
        &coords.map(|v| v as f32),
        &bounds,
        basin,
    );
    println!(
        "Bounds energy: {:.6}, energy/atom: {:.6}",
        e_total,
        e_total / n as f32
    );

    // Step 5: Second minimization if 4D
    if use_4d {
        let mut need_more2 = 1;
        let mut iters2 = 0;
        while need_more2 != 0 {
            need_more2 = minimize_bfgs_rdkit(
                &mut coords,
                &bounds,
                &chiral_sets,
                200,
                force_tol,
                basin,
                1.0,
                0.2,
            );
            iters2 += 1;
        }
        println!("\n=== After Bounds FF (stage 2, {} rounds) ===", iters2);
        for i in 0..n.min(5) {
            println!(
                "  {}: ({:.8}, {:.8}, {:.8})",
                i,
                coords[(i, 0)],
                coords[(i, 1)],
                coords[(i, 2)]
            );
        }
    }

    // Step 6: Drop to 3D
    let coords3d = coords.columns(0, 3).into_owned();
    println!("\n=== After Drop to 3D (first 5) ===");
    for i in 0..n.min(5) {
        println!(
            "  {}: ({:.8}, {:.8}, {:.8})",
            i,
            coords3d[(i, 0)],
            coords3d[(i, 1)],
            coords3d[(i, 2)]
        );
    }

    // Step 7: ETKDG 3D FF minimization
    let ff = build_etkdg_3d_ff_with_torsions(&mol, &coords3d, &bounds, &[]);
    let refined = minimize_etkdg_3d_bfgs(&mol, &coords3d, &ff, 300, 1e-3);
    println!("\n=== After ETKDG 3D Minimization (first 5) ===");
    for i in 0..n.min(5) {
        println!(
            "  {}: ({:.8}, {:.8}, {:.8})",
            i,
            refined[(i, 0)],
            refined[(i, 1)],
            refined[(i, 2)]
        );
    }

    // Load RDKit final coordinates for comparison
    let rdkit_coords = load_npy_f64("/tmp/rdkit_coords_fail.npy");
    println!("\n=== RDKit Final Coordinates (first 5) ===");
    for i in 0..n.min(5) {
        println!(
            "  {}: ({:.8}, {:.8}, {:.8})",
            i,
            rdkit_coords[(i, 0)],
            rdkit_coords[(i, 1)],
            rdkit_coords[(i, 2)]
        );
    }

    // Compute RMSD (with alignment)
    let our_f32 = refined.map(|v| v as f32);
    let rdkit_f32 = rdkit_coords.map(|v| v as f32);
    // Simple centroid-aligned RMSD
    let mut our_cx = 0.0f32;
    let mut our_cy = 0.0f32;
    let mut our_cz = 0.0f32;
    let mut rd_cx = 0.0f32;
    let mut rd_cy = 0.0f32;
    let mut rd_cz = 0.0f32;
    for i in 0..n {
        our_cx += our_f32[(i, 0)];
        our_cy += our_f32[(i, 1)];
        our_cz += our_f32[(i, 2)];
        rd_cx += rdkit_f32[(i, 0)];
        rd_cy += rdkit_f32[(i, 1)];
        rd_cz += rdkit_f32[(i, 2)];
    }
    our_cx /= n as f32;
    our_cy /= n as f32;
    our_cz /= n as f32;
    rd_cx /= n as f32;
    rd_cy /= n as f32;
    rd_cz /= n as f32;
    let mut sum_sq = 0.0f32;
    for i in 0..n {
        let dx = (our_f32[(i, 0)] - our_cx) - (rdkit_f32[(i, 0)] - rd_cx);
        let dy = (our_f32[(i, 1)] - our_cy) - (rdkit_f32[(i, 1)] - rd_cy);
        let dz = (our_f32[(i, 2)] - our_cz) - (rdkit_f32[(i, 2)] - rd_cz);
        sum_sq += dx * dx + dy * dy + dz * dz;
    }
    let rmsd_no_rotation = (sum_sq / n as f32).sqrt();
    println!(
        "\nRMSD (centroid-only, no rotation): {:.4}",
        rmsd_no_rotation
    );
}