Add option for GNPS index multi-charge scoring and PIN the version

bin/scripts/index_all_pairwise.py

@@ -192,7 +192,7 @@ def find_bin_range(entries, target_bin):
 
 
 @numba.njit
-def compute_all_pairs(spectra, shared_entries, shifted_entries, tolerance, threshold):
+def compute_all_pairs(spectra, shared_entries, shifted_entries, tolerance, threshold, scoring_func):
     results = List()
     n_spectra = len(spectra)
 
@@ -241,7 +241,7 @@ def compute_all_pairs(spectra, shared_entries, shifted_entries, tolerance, thres
         exact_matches = List()
         for spec_idx, _ in candidates[:TOPPRODUCTS * 2]:
             target_spec = spectra[spec_idx]
-            score, shared, shifted = calculate_exact_score_GNPS(spectra[query_idx], target_spec,tolerance)
+            score, shared, shifted = scoring_func(spectra[query_idx], target_spec,tolerance)
             if score >= threshold:
                 exact_matches.append((spec_idx, score, shared, shifted))
 
@@ -251,9 +251,105 @@ def compute_all_pairs(spectra, shared_entries, shifted_entries, tolerance, thres
 
     return results
 
+@numba.njit(fastmath=True)
+def calculate_exact_score_GNPS(query_spec, target_spec, TOLERANCE):
+    """Numba-optimized cosine scoring with shift handling"""
+    q_mz = query_spec[0]
+    q_int = query_spec[1]
+    q_pre = query_spec[2]
+    q_charge = query_spec[3]
+
+    t_mz = target_spec[0]
+    t_int = target_spec[1]
+    t_pre = target_spec[2]
+
+    # Calculate precursor mass difference (assuming charge=1)
+    precursor_mass_diff = (q_pre - t_pre)*q_charge
+    allow_shift = True
+    fragment_tol = TOLERANCE
+
+    # Pre-allocate arrays for matches (adjust size as needed)
+    max_matches = len(q_mz) * 2  # Estimate maximum possible matches
+    scores_arr = np.zeros(max_matches, dtype=np.float32)
+    idx_q = np.zeros(max_matches, dtype=np.int32)
+    idx_t = np.zeros(max_matches, dtype=np.int32)
+    match_count = 0
+
+    # For each peak in query spectrum
+    for q_idx in range(len(q_mz)):
+        q_mz_val = q_mz[q_idx]
+        q_int_val = q_int[q_idx]
+
+        # For each possible shift (charge=1)
+        num_shifts = 1
+        if allow_shift and abs(precursor_mass_diff) >= fragment_tol:
+            num_shifts += 1
+
+        for shift_idx in range(num_shifts):
+            if shift_idx == 0:
+                # No shift
+                adjusted_mz = q_mz_val
+            else:
+                # Apply shift
+                adjusted_mz = q_mz_val - precursor_mass_diff
+
+            # Find matching peaks in target using binary search
+            start = np.searchsorted(t_mz, adjusted_mz - fragment_tol)
+            end = np.searchsorted(t_mz, adjusted_mz + fragment_tol)
+
+            for t_idx in range(start, end):
+                if match_count >= max_matches:
+                    break  # Prevent overflow
+                if abs(t_mz[t_idx] - adjusted_mz)<fragment_tol:
+                    # Store match information
+                    scores_arr[match_count] = q_int_val * t_int[t_idx]
+                    idx_q[match_count] = q_idx
+                    idx_t[match_count] = t_idx
+                    match_count += 1
+
+    # Sort matches by score descending using argsort
+    if match_count == 0:
+        return 0.0, 0.0, 0.0
+
+    # Get valid matches
+    valid_scores = scores_arr[:match_count]
+    valid_q_idx = idx_q[:match_count]
+    valid_t_idx = idx_t[:match_count]
+
+    # Argsort descending
+    sort_order = np.argsort(-valid_scores)
+
+    # Track used peaks
+    q_used = np.zeros(len(q_mz), dtype=np.bool_)
+    t_used = np.zeros(len(t_mz), dtype=np.bool_)
+    total = 0.0
+    shared = 0.0
+    shifted = 0.0
+
+    # Accumulate top matches
+    for i in sort_order:
+        q_idx = valid_q_idx[i]
+        t_idx = valid_t_idx[i]
+
+        if not q_used[q_idx] and not t_used[t_idx]:
+            score = valid_scores[i]
+            total += score
+
+            # Determine match type
+            mz_diff = abs(q_mz[q_idx] - t_mz[t_idx])
+            if mz_diff <= fragment_tol:
+                shared += score
+            else:
+                shifted += score
+
+            q_used[q_idx] = True
+            t_used[t_idx] = True
+
+    return total, shared, shifted
+
 
 @numba.njit
-def calculate_exact_score_GNPS(query_spec, target_spec, TOLERANCE):
+def calculate_exact_score_GNPS_multi_charge(query_spec, target_spec, TOLERANCE):
     """Numba-optimized cosine scoring with shift handling based on original GNPS code."""
     q_mz = query_spec[0]
     q_int = query_spec[1]
@@ -445,6 +541,8 @@ def parse_arguments():
                         help="Similarity score threshold")
     parser.add_argument("--threads", type=int, default=1,
                         help="Number of processing threads")
+    parser.add_argument("--multi_charge", action="store_true",
+                        help="Use GNPS multi-charge scoring")
     return parser.parse_args()
 
 
@@ -470,10 +568,14 @@ def main():
     shared_idx = create_index(numba_spectra, False, args.tolerance, SHIFTED_OFFSET)
     shifted_idx = create_index(numba_spectra, True, args.tolerance, SHIFTED_OFFSET)
 
+    if args.multi_charge:
+        scoring_func = calculate_exact_score_GNPS_multi_charge
+    else:
+        scoring_func = calculate_exact_score_GNPS
     # Compute matches
     print("Computing pairs...")
     matches = compute_all_pairs(numba_spectra, shared_idx, shifted_idx,
-                                args.tolerance, args.threshold)
+                                args.tolerance, args.threshold, scoring_func)
 
     # Write output in Nextflow-compatible format
     print("Writing results...")

nf_workflow.nf

@@ -26,7 +26,8 @@ params.precursor_filter = "1"
 
 // Molecular Networking Options
 params.topology = "classic" // or can be transitive
-params.cal_all_pairs ='gnps' //or can be index
+params.cal_all_pairs ='gnps' //or can be index_single and index_multi
+
 
 params.parallelism = 24
 params.networking_min_matched_peaks = 6
@@ -250,7 +251,7 @@ process calculatePairs {
     """
 }
 
-process calculatePairs_index {
+process calculatePairs_index_single {
     publishDir "$params.publishdir/nf_output/temp_pairs", mode: 'copy'
 
     conda "$TOOL_FOLDER/conda_env.yml"
@@ -269,6 +270,27 @@ process calculatePairs_index {
     --threshold $params.networking_min_cosine
     """
 }
+
+process calculatePairs_index_multi {
+    publishDir "$params.publishdir/nf_output/temp_pairs", mode: 'copy'
+
+    conda "$TOOL_FOLDER/conda_env.yml"
+
+    input:
+    file spectrum_file
+
+    output:
+    file "*_aligns.tsv" optional true
+
+    """
+    python $TOOL_FOLDER/scripts/index_all_pairwise.py \
+    -t $spectrum_file \
+    -o 0.params_aligns.tsv\
+    --tolerance $params.fragment_tolerance \
+    --threshold $params.networking_min_cosine \
+    --multi_charge
+    """
+}
 // Creating the metadata file
 process createMetadataFile {
     publishDir "$params.publishdir/nf_output/metadata", mode: 'copy'
@@ -601,8 +623,11 @@ workflow {
         params_ch = networkingGNPSPrepParams(clustered_spectra_ch)
         networking_results_temp_ch = calculatePairs(clustered_spectra_ch, params_ch.collect())
     }
-    else if (params.cal_all_pairs == "index"){
-        networking_results_temp_ch = calculatePairs_index(clustered_spectra_ch)
+    else if (params.cal_all_pairs == "index_single"){
+        networking_results_temp_ch = calculatePairs_index_single(clustered_spectra_ch)
+    }
+    else if(params.cal_all_pairs == "index_multi"){
+        networking_results_temp_ch = calculatePairs_index_multi(clustered_spectra_ch)
     }
     merged_networking_pairs_ch = networking_results_temp_ch.collectFile(name: "merged_pairs.tsv", storeDir: "./nf_output/networking", keepHeader: true)
 

workflowinput.yaml

@@ -1,7 +1,7 @@
 workflowname: classical_networking_workflow
 workflowdescription: This is Classical Molecular Networking for GNPS2
 workflowlongdescription: This is Classical Molecular Networking for GNPS2
-workflowversion: "2025.04.11"
+workflowversion: "2025.05.06"
 workflowfile: nf_workflow.nf
 workflowautohide: false
 adminonly: false
@@ -160,8 +160,10 @@ parameterlist:
       options:
         - value: "gnps"
           display: "GNPS"
-        - value: "index"
-          display: "GNPS_Index1"
+        - value: "index_single"
+          display: "GNPS_Index1_single_charge"
+        - value: "index_multi"
+          display: "GNPS_Index1_multi_charge"
 
     - displayname: Network Topology Parameters
       paramtype: section