Gene and protein sets for k-mer comparisons
Each gene and each complete exact protein sequence appears in one group only.
| Group | Genes | Exact protein sequences |
|---|---|---|
| Normal | 18,064 | 126,691 |
| Unassessed | 610 | 1,089 |
| Reproductive | 1,022 | 1,782 |
| Other | 401 | 423 |
Choose a download
| Group | Gene list | Protein list | Sequences |
|---|---|---|---|
| Normal | Genes | Proteins | FASTA |
| Unassessed | Genes | Proteins | FASTA |
| Reproductive | Genes | Proteins | FASTA |
| Other | Genes | Proteins | FASTA |
How shared proteins are assigned
All complete annotated isoforms are compared by exact amino-acid sequence. If any encoding gene is Normal, the shared protein is Normal. Otherwise, any Unassessed gene makes it Unassessed; then Reproductive takes precedence over Other. Every original gene, transcript and protein record remains in the occurrence table. Verified alternate-contig identities use the existing annotation-aware resolver; matching names alone do not merge loci.
Gene groups follow the approved expression screen. External gene-specific normal-tissue evidence adds MAGEA12 to Normal; its original HPA expression group remains visible. These are sequence-defined protein forms, without PTM-specific distinctions. All isoforms inherit gene-level evidence; isoform-specific expression has not been established.
Build a k-mer exclusion background
- Enumerate target k-mers from Reproductive.
- Subtract every k-mer in Normal, Unassessed, and the additional background FASTA containing incomplete or unusual translations from Normal/Unassessed genes.
- Keep record boundaries intact. For incomplete or unusual translations, enumerate only contiguous standard amino-acid windows; never join across an ambiguous residue or a stop.
- Other is available separately. It contains Disposable/Silent below-threshold expression, and does not supply positive reproductive evidence.
Full-protein sets are disjoint; their k-mers can still overlap. The subtraction step remains necessary. The machine-readable recipe lists the exact files. Complete proteins with any observed normal expression are an optional stricter background; this can overlap Reproductive because a passing RNA cutoff does not erase a measured trace signal. For incomplete translations, the review table preserves each gene's group; the optional complete-protein file does not establish exhaustive sequence coverage.
Coverage and provenance
Pinned human Ensembl 115 / GRCh38.p14, all peptide FASTA records, with transcript completeness checked against start/stop codons and CDS completeness flags in the GTF. The gene categories use HPA v23. No expression cutoffs are transferred to another species.
34,361 translation records are incomplete or contain nonstandard/ambiguous residues. They remain listed with reasons and gene groups. Only records from Normal/Unassessed genes enter the additional background; incompleteness alone does not turn a Reproductive gene into a normal-tissue exclusion. 307 classified genes have no complete usable protein; that is a coverage gap, not evidence of sequence absence.
- All gene assignments · All protein assignments · Counts
- Gene mapping decisions · Reference identity · Checksums
- CTA curation report · Expression groups
Rebuild with python scripts/normal_sequence_sets_report.py --out <empty-directory> --categories <expression-report> --gtf <Ensembl-115-GTF.gz> --proteins <Ensembl-115-peptide-FASTA.gz>.