diff --git a/pyensembl/database.py b/pyensembl/database.py
index 70a5435..571963c 100644
--- a/pyensembl/database.py
+++ b/pyensembl/database.py
@@ -16,7 +16,7 @@
 
 import datacache
 from typechecks import require_integer, require_string
-from gtfparse import read_gtf, create_missing_features
+from gtfparse import GENCODE_BIOTYPE_ALIASES, create_missing_features, read_gtf
 
 from .common import memoize
 from .normalization import normalize_chromosome, normalize_strand
@@ -626,6 +626,11 @@ def _load_gtf_as_dataframe(self, usecols=None, features=None):
                 "seqname": normalize_chromosome,
                 "strand": normalize_strand,
             },
+            # GENCODE GTFs use gene_type / transcript_type where Ensembl uses
+            # gene_biotype / transcript_biotype. attribute_aliases (gtfparse
+            # 2.7.0+) renames the GENCODE columns onto the Ensembl names at
+            # parse time so the rest of pyensembl can stay Ensembl-centric.
+            attribute_aliases=GENCODE_BIOTYPE_ALIASES,
             infer_biotype_column=True,
             usecols=usecols,
             features=features,
diff --git a/pyensembl/version.py b/pyensembl/version.py
index cbd495c..af9817b 100644
--- a/pyensembl/version.py
+++ b/pyensembl/version.py
@@ -1,4 +1,4 @@
-__version__ = "2.9.7"
+__version__ = "2.9.8"
 
 def print_version():
     print(f"v{__version__}")
diff --git a/pyproject.toml b/pyproject.toml
index 2e32240..1a2f358 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -25,7 +25,7 @@ dependencies = [
     "datacache>=1.4.0,<2.0.0",
     "memoized-property>=1.0.2",
     "tinytimer>=0.0.0,<1.0.0",
-    "gtfparse>=2.6.0,<3.0.0",
+    "gtfparse>=2.7.0,<3.0.0",
     "serializable>=0.2.1,<2.0.0",
     "numpy>=2.0.0,<3.0.0",
 ]
diff --git a/tests/test_gencode_biotype_aliases.py b/tests/test_gencode_biotype_aliases.py
new file mode 100644
index 0000000..18e2102
--- /dev/null
+++ b/tests/test_gencode_biotype_aliases.py
@@ -0,0 +1,104 @@
+"""
+Issue #335 part 1: pyensembl now passes gtfparse's GENCODE_BIOTYPE_ALIASES
+into read_gtf, so a vanilla GENCODE GTF (which carries gene_type /
+transcript_type) gets normalized onto the Ensembl gene_biotype /
+transcript_biotype column names at parse time. Without this, GENCODE
+genomes silently looked like every transcript had no biotype and
+Transcript.is_protein_coding returned False for everything (see #335
+issue body).
+"""
+
+import os
+
+from pyensembl import Genome
+
+from .common import TemporaryDirectory, eq_
+
+
+GENCODE_STYLE_GTF = """\
+1\tHAVANA\tgene\t100\t800\t.\t+\t.\tgene_id "ENSGTEST00000000001.4"; gene_type "protein_coding"; gene_name "FAKE1";
+1\tHAVANA\ttranscript\t100\t800\t.\t+\t.\tgene_id "ENSGTEST00000000001.4"; transcript_id "ENSTTEST00000000001.5"; gene_type "protein_coding"; transcript_type "protein_coding"; gene_name "FAKE1";
+1\tHAVANA\texon\t100\t800\t.\t+\t.\tgene_id "ENSGTEST00000000001.4"; transcript_id "ENSTTEST00000000001.5"; exon_number "1"; exon_id "ENSETEST00000000001.2"; gene_type "protein_coding"; transcript_type "protein_coding"; gene_name "FAKE1";
+1\tHAVANA\tCDS\t100\t798\t.\t+\t0\tgene_id "ENSGTEST00000000001.4"; transcript_id "ENSTTEST00000000001.5"; exon_number "1"; exon_id "ENSETEST00000000001.2"; protein_id "ENSPTEST00000000001.3"; gene_type "protein_coding"; transcript_type "protein_coding"; gene_name "FAKE1";
+1\tHAVANA\tstart_codon\t100\t102\t.\t+\t0\tgene_id "ENSGTEST00000000001.4"; transcript_id "ENSTTEST00000000001.5"; gene_type "protein_coding"; transcript_type "protein_coding"; gene_name "FAKE1";
+1\tHAVANA\tstop_codon\t799\t801\t.\t+\t0\tgene_id "ENSGTEST00000000001.4"; transcript_id "ENSTTEST00000000001.5"; gene_type "protein_coding"; transcript_type "protein_coding"; gene_name "FAKE1";
+1\tHAVANA\tgene\t2000\t3000\t.\t+\t.\tgene_id "ENSGTEST00000000002.1"; gene_type "lncRNA"; gene_name "FAKE2";
+1\tHAVANA\ttranscript\t2000\t3000\t.\t+\t.\tgene_id "ENSGTEST00000000002.1"; transcript_id "ENSTTEST00000000002.1"; gene_type "lncRNA"; transcript_type "lncRNA"; gene_name "FAKE2";
+1\tHAVANA\texon\t2000\t3000\t.\t+\t.\tgene_id "ENSGTEST00000000002.1"; transcript_id "ENSTTEST00000000002.1"; exon_number "1"; exon_id "ENSETEST00000000002.1"; gene_type "lncRNA"; transcript_type "lncRNA"; gene_name "FAKE2";
+"""
+
+
+def _make_genome(tmpdir):
+    gtf_path = os.path.join(tmpdir, "gencode_style.gtf")
+    with open(gtf_path, "w") as f:
+        f.write(GENCODE_STYLE_GTF)
+    return Genome(
+        reference_name="GRCh38",
+        annotation_name="_test_gencode_biotype_aliases_335p1",
+        gtf_path_or_url=gtf_path,
+        cache_directory_path=tmpdir,
+    )
+
+
+def test_gencode_gene_type_normalized_to_gene_biotype():
+    """A GENCODE GTF (with gene_type / transcript_type attributes) is
+    parsed so that the sqlite database stores gene_biotype /
+    transcript_biotype, and querying by Ensembl biotype names works."""
+    with TemporaryDirectory() as tmpdir:
+        genome = _make_genome(tmpdir)
+        genome.index()
+
+        # Biotype columns exist in the gene + transcript tables
+        assert genome.db.column_exists("gene", "gene_biotype")
+        assert genome.db.column_exists("transcript", "transcript_biotype")
+
+        # Filter by biotype works against the renamed columns
+        coding_genes = genome.genes(biotype="protein_coding")
+        eq_(len(coding_genes), 1)
+        eq_(coding_genes[0].id, "ENSGTEST00000000001.4")
+        eq_(coding_genes[0].biotype, "protein_coding")
+
+        ncrna_genes = genome.genes(biotype="lncRNA")
+        eq_(len(ncrna_genes), 1)
+        eq_(ncrna_genes[0].id, "ENSGTEST00000000002.1")
+
+
+def test_gencode_transcript_type_normalized_to_transcript_biotype():
+    with TemporaryDirectory() as tmpdir:
+        genome = _make_genome(tmpdir)
+        genome.index()
+
+        coding = genome.transcripts(biotype="protein_coding")
+        eq_(len(coding), 1)
+        eq_(coding[0].id, "ENSTTEST00000000001.5")
+        eq_(coding[0].biotype, "protein_coding")
+
+
+def test_gencode_transcript_is_protein_coding_true_after_alias():
+    """The original #335 repro: Variant(...).effects() returned
+    NoncodingTranscript for every transcript because is_protein_coding
+    was False. With attribute_aliases the biotype column lands on the
+    Ensembl name and is_protein_coding returns True for coding rows."""
+    with TemporaryDirectory() as tmpdir:
+        genome = _make_genome(tmpdir)
+        genome.index()
+        coding = genome.transcript_by_id("ENSTTEST00000000001.5")
+        assert coding.is_protein_coding is True
+        noncoding = genome.transcript_by_id("ENSTTEST00000000002.1")
+        assert noncoding.is_protein_coding is False
+
+
+def test_pure_ensembl_gtf_unaffected_by_aliasing():
+    """Vanilla Ensembl GTFs already use gene_biotype / transcript_biotype.
+    Passing the alias dict must be a no-op for them (gtfparse skips
+    missing alias sources). Reuses the existing mouse partial fixture
+    which is Ensembl 81 style."""
+    from .data import (
+        custom_mouse_genome_grcm38_subset,
+        setup_init_custom_mouse_genome,
+    )
+    setup_init_custom_mouse_genome()
+    coding = custom_mouse_genome_grcm38_subset.transcripts(biotype="protein_coding")
+    eq_(len(coding), 1)
+    eq_(coding[0].id, "ENSMUST00000103109")
+    assert coding[0].is_protein_coding is True