min-contig for train #24; fix mem ab initio limit #22; --skip-predictors #22; bump version

Author: Jon Palmer

CITATION.cff

@@ -1,4 +1,4 @@
-cff-version: version = "25.7.16"
+cff-version: version = "25.7.19"
 title: 'funannotate2: eukaryotic genome annotation'
 message: >-
   If you use this software, please cite it using the
@@ -17,5 +17,5 @@ keywords:
   - functional annotation
   - consensus gene models
 license: BSD-2-Clause
-version: version = "25.7.16"
-date-released: '2025-07-16'
+version: version = "25.7.19"
+date-released: '2025-07-19'

funannotate2/__main__.py

@@ -286,6 +286,13 @@ def predict_subparser(subparsers):
         help="Memory limit in GB to adjust CPU allocation",
         metavar="",
     )
+    optional_args.add_argument(
+        "--skip-predictors",
+        nargs="+",
+        choices=["snap", "augustus", "glimmerhmm", "genemark"],
+        help="Skip specific ab initio predictors (choices: snap, augustus, glimmerhmm, genemark)",
+        metavar="",
+    )
     other_args = group.add_argument_group("Other arguments")
     other_args.add_argument(
         "-h",
@@ -349,6 +356,14 @@ def train_subparser(subparsers):
         help="Max length for fasta headers",
         metavar="",
     )
+    optional_args.add_argument(
+        "--min-contig-length",
+        default=10000,
+        dest="min_contig_length",
+        type=int,
+        help="Minimum contig length to use for training (default: 10000)",
+        metavar="",
+    )
     optional_args.add_argument(
         "--busco-lineage",
         dest="busco_lineage",

funannotate2/fastx.py

@@ -259,6 +259,54 @@ def simplify_headers_drop(inputfile, keepfile, dropfile, base="contig_", drop=[]
     return names
 
 
+def filter_contigs_by_length(inputfile, outputfile, min_length=10000, base="contig_"):
+    """
+    Filter contigs by minimum length and simplify headers.
+
+    This function reads a FASTA file, filters contigs based on a minimum length threshold,
+    and writes the filtered contigs to an output file with simplified headers. This is
+    particularly useful for training ab initio gene predictors where short contigs
+    are not informative and can hurt training quality.
+
+    Args:
+        inputfile (str): Path to the input FASTA file.
+        outputfile (str): Path to the output file for filtered contigs.
+        min_length (int, optional): Minimum contig length to keep. Defaults to 10000.
+        base (str, optional): Base string for simplified headers. Defaults to "contig_".
+
+    Returns:
+        tuple: A tuple containing:
+            - dict: Mapping of simplified headers to original headers for kept contigs.
+            - int: Number of contigs kept.
+            - int: Number of contigs filtered out.
+            - int: Total length of kept contigs.
+            - int: Total length of filtered contigs.
+    """
+    names = {}
+    kept_count = 0
+    filtered_count = 0
+    kept_length = 0
+    filtered_length = 0
+
+    with open(outputfile, "w") as outfile:
+        for title, seq in pyfastx.Fasta(inputfile, build_index=False):
+            seq_length = len(seq)
+
+            if seq_length >= min_length:
+                # Keep this contig
+                kept_count += 1
+                kept_length += seq_length
+                simplified_name = f"{base}{kept_count}"
+                names[simplified_name] = title
+                outfile.write(f">{simplified_name}\n{softwrap(seq)}\n")
+            else:
+                # Filter out this contig
+                filtered_count += 1
+                filtered_length += seq_length
+
+    return names, kept_count, filtered_count, kept_length, filtered_length
+
+
 def list2groups(L):
     """
     Identify groups of continuous numbers in a list.