BioMart Pig ↔ Mouse 1:1 Ortholog Mapping for Cross-Species Proteomics (R + Python Tutorial)
Step-by-step tutorial: download the Ensembl BioMart Release 111 Sus scrofa ↔ Mus musculus 1:1 ortholog table and join it to your per-species protein quantification in R (biomaRt) and Python (pandas). Why gene-symbol matching is fragile, how to handle one-to-many and many-to-many orthologs, and how to validate your mapping against UniProt.
Why Gene-Symbol Matching Isn't Good Enough
If you ran cross-species LC-MS/MS proteomics — Porcine ECM vs Mouse Matrigel, human-vs-mouse cells, host-pathogen — you eventually need to join two species' protein tables to compare. The fast-and-loose way is gene symbol matching with uppercase normalization:
df['GENE_UPPER'] = df['Genes'].str.upper()
merged = porcine_df.merge(mouse_df, on='GENE_UPPER')
This works for most well-annotated proteins. It fails silently for:
- Paralog families: a symbol match picks whichever family member shares the name, without checking that it is the true ortholog
- Symbol drift: gene symbols change between annotations; mouse
Hba-a1vs humanHBA1look identical but aren't always 1:1 orthologs - Many-to-many: some gene families (immunoglobulins, MHC, olfactory receptors) have lineage-specific expansions
- Case where one species has a gene and the other doesn't
The robust alternative is to use the Ensembl BioMart 1:1 ortholog table — Ensembl Compara's curated set of orthologs flagged as "one-to-one" between species pairs. This guide shows the exact workflow for Sus scrofa ↔ Mus musculus at Ensembl Release 111 (January 2024, the release this project pinned), with R and Python implementations, and the numbers I got when I ran it against 111 and the current release, 116.
Related: Why You Must NOT Merge Species FASTA Databases in Cross-Species Proteomics covers the upstream principle — why you have two separate species tables that need joining in the first place.
Step 1 — What Is "1:1 Ortholog" in BioMart
Ensembl Compara classifies cross-species gene relationships:
| Type | Meaning |
|---|---|
ortholog_one2one | One gene in each species, single common ancestor |
ortholog_one2many | One gene → multiple copies in the other species (lineage expansion) |
ortholog_many2many | Multiple in both (gene family expansion in both lineages) |
paralog | Within-species duplication |
When I queried pig → mouse in releases 111 and 116, only the three ortholog_* types came back. For cross-species quantitative proteomics comparison, only ortholog_one2one is safe. The others require case-by-case handling.
Step 2 — Download via BioMart Web UI
The fastest path:
- Go to https://www.ensembl.org/biomart/martview. In October 2026 this redirects to
jun2026.archive.ensembl.org(Ensembl 116). For release 111 usehttps://jan2024.archive.ensembl.org/biomart/martview - CHOOSE DATABASE: Ensembl Genes 111 (or the release you pin)
- CHOOSE DATASET: Pig genes (Sus scrofa)
- Filters: none
- Attributes:
- Features → GENE →
Gene stable ID,Gene name(UniProt IDs are on a different attribute page; BioMart refuses to mix pages in one query, see Step 3) - HOMOLOGUES → Orthologues → expand "Mouse Orthologues":
Mouse gene stable IDMouse gene nameMouse homology type(this is where you'll filter toortholog_one2one)% identity(optional, useful for QC)
- Features → GENE →
- Results → Compressed: No → Format: TSV → Go
Save as biomart_pig_mouse_r111.tsv. In release 111 I got 26,676 rows with a mouse homolog: 16,510 one-to-one, 3,224 one-to-many and 6,942 many-to-many.
Filter to 1:1 in your downstream code:
import pandas as pd
bm = pd.read_csv('biomart_pig_mouse_r111.tsv', sep='\t')
bm_one2one = bm[bm['Mouse homology type'] == 'ortholog_one2one'].copy()
print(f"1:1 orthologs: {len(bm_one2one):,}") # 16,510 for release 111
For the Pig (Sus scrofa) ↔ Mouse (Mus musculus) pair I counted 16,510 one-to-one pairs in release 111 and 16,529 in release 116 (queried 10 October 2026). 16,004 pairs were in both; 506 were only in 111 and 525 only in 116, so about 3% of the table changed between the two releases. [corrected 2026-10: an earlier version said "roughly 13,476" without a source]
Step 3 — Programmatic Download (R, biomaRt)
If you want the download reproducible in a script:
With an up-to-date biomaRt the usual call is useEnsembl(biomart = "genes", dataset = "sscrofa_gene_ensembl", version = 111) followed by getBM(). In my run (biomaRt 2.58.2, R 4.3), that failed with invalid 'file' argument while biomaRt listed the Ensembl archives, and so did useMart(..., host = "https://jan2024.archive.ensembl.org"); www.ensembl.org/biomart now answers with an HTTP 308 redirect, which older biomaRt does not follow. If your biomaRt works, keep using it. What worked for me in plain R, with no package, is the same XML query the BioMart web UI sends:
# Pig -> mouse homologues from Ensembl 111 (January 2024 archive), plain R
attrs <- c("ensembl_gene_id", "external_gene_name", "mmusculus_homolog_ensembl_gene",
"mmusculus_homolog_associated_gene_name", "mmusculus_homolog_orthology_type",
"mmusculus_homolog_perc_id")
xml <- paste0('<?xml version="1.0" encoding="UTF-8"?><!DOCTYPE Query>',
'<Query virtualSchemaName="default" formatter="TSV" header="0" uniqueRows="1">',
'<Dataset name="sscrofa_gene_ensembl" interface="default">',
paste0('<Attribute name="', attrs, '"/>', collapse = ""), '</Dataset></Query>')
url <- paste0("https://jan2024.archive.ensembl.org/biomart/martservice?query=",
URLencode(xml, reserved = TRUE))
orthos <- read.delim(url, header = FALSE, col.names = attrs, colClasses = "character")
orthos <- orthos[orthos$mmusculus_homolog_ensembl_gene != "", ]
# Filter to 1:1 (16,510 rows for release 111 in my run)
one2one <- subset(orthos, mmusculus_homolog_orthology_type == "ortholog_one2one")
write.table(one2one, "biomart_pig_mouse_one2one_r111.tsv",
sep = "\t", row.names = FALSE, quote = FALSE)
Keep homologue attributes on their own: adding a Features-page attribute such as uniprotswissprot to the same query makes BioMart reply Query ERROR: caught BioMart::Exception::Usage: Attributes from multiple attribute pages are not allowed. Fetch UniProt IDs in a second query (ensembl_gene_id + uniprotswissprot) and merge by gene ID.
Caveats:
- Record the release number, the host you queried and the download date in your Methods.
- Pin the release for a paper; don't rely on "current".
- BioMart's own error messages arrive as plain text instead of a table (for example
Query ERROR: caught BioMart::Exception::Usage: Attributes from multiple attribute pages are not allowed). Check the first line of what you saved.
Step 4 — Programmatic Download (Python, pybiomart)
from pybiomart import Server
server = Server(host='http://jan2024.archive.ensembl.org') # release 111. Use http://: pybiomart
# adds port 80, and an https:// host failed
mart = server['ENSEMBL_MART_ENSEMBL']
dataset = mart['sscrofa_gene_ensembl']
orthos = dataset.query(attributes=[
'ensembl_gene_id',
'external_gene_name',
'mmusculus_homolog_ensembl_gene',
'mmusculus_homolog_associated_gene_name',
'mmusculus_homolog_orthology_type',
'mmusculus_homolog_perc_id',
], use_attr_names=True) # keep attribute names as columns; the default returns display
# names ('Gene name', 'Mouse homology type', ...)
one2one = orthos[orthos['mmusculus_homolog_orthology_type'] == 'ortholog_one2one']
one2one.to_csv('biomart_pig_mouse_one2one_r111.tsv', sep='\t', index=False)
print(f"Wrote {len(one2one):,} 1:1 orthologs")
pybiomart 0.2.0 still worked in my test (October 2026), but its last release was in May 2017 (PyPI). Alternatives:
- A direct query to the
martserviceendpoint with the same XML BioMart's web UI generates (no package needed; this is how I counted the pairs above) - The Ensembl REST API,
https://rest.ensembl.org/homology/...(per-gene, slower for whole genomes)
Step 5 — Join to Your Proteomics Quantification Tables
After species-separated MaxQuant/DIA-NN searches, you have:
porcine_pg_matrix.tsv(Porcine proteins, columns: Genes, sample_1, sample_2, ...)mouse_pg_matrix.tsv(Mouse proteins, columns: Genes, sample_1, sample_2, ...)
Join via the BioMart table:
import pandas as pd
porcine = pd.read_csv('porcine_pg_matrix.tsv', sep='\t')
mouse = pd.read_csv('mouse_pg_matrix.tsv', sep='\t')
bm = pd.read_csv('biomart_pig_mouse_one2one_r111.tsv', sep='\t')
# Keep only essential ortholog mapping columns
bm_map = bm[['external_gene_name', 'mmusculus_homolog_associated_gene_name']].rename(
columns={
'external_gene_name': 'pig_gene',
'mmusculus_homolog_associated_gene_name': 'mouse_gene',
}
)
bm_map = bm_map.dropna()
# Add an ortholog ID
bm_map['ortho_id'] = range(len(bm_map))
# Map porcine proteins to ortholog IDs
porcine_mapped = porcine.merge(
bm_map[['pig_gene', 'ortho_id']],
left_on='Genes', right_on='pig_gene', how='inner'
)
# Map mouse proteins to ortholog IDs
mouse_mapped = mouse.merge(
bm_map[['mouse_gene', 'ortho_id']],
left_on='Genes', right_on='mouse_gene', how='inner'
)
# Inner-join on ortho_id — now porcine and mouse rows are aligned by orthology
combined = porcine_mapped.merge(
mouse_mapped, on='ortho_id', suffixes=('_pig', '_mouse')
)
print(f"Porcine proteins quantified: {len(porcine):,}")
print(f"Mouse proteins quantified: {len(mouse):,}")
print(f"After 1:1 ortholog join: {len(combined):,} ortholog pairs")
What to expect (realistic numbers from PRIDE PXD023694-style data)
- Porcine intestine ECM after QC: ~1,000 proteins
- Mouse Matrigel after QC: ~700 proteins
- 1:1 ortholog pairs after join: ~280-300
Why so few? Two compounding effects:
- Not every protein is detected in both species (biological + technical reasons)
- Some proteins lack 1:1 orthologs (paralog expansions, lineage-specific genes)
A low rate is not unusual when one species' identifications include unnamed or lineage-specific genes. Look at which proteins dropped out rather than aiming for a particular percentage.
Step 6 — Validate Your Mapping
After joining, sanity-check before running statistics:
Check 1 — known orthologs are present
known_orthos = [
('COL1A1', 'Col1a1'), # collagen I alpha 1
('LAMB1', 'Lamb1'), # laminin beta 1
('GAPDH', 'Gapdh'), # housekeeping
('ACTB', 'Actb'), # actin
('FN1', 'Fn1'), # fibronectin
]
for pig, mouse in known_orthos:
found = ((bm_map['pig_gene'] == pig) & (bm_map['mouse_gene'] == mouse)).any()
print(f" {pig} ↔ {mouse}: {'✓' if found else '✗ MISSING'}")
If any of these are missing, recheck your BioMart download (wrong filter? wrong release? incomplete columns?).
Check 2 — duplicate handling
A "1:1" ortholog table should have at most one row per pig gene and per mouse gene. Verify:
print("Pig genes appearing >1x:", (bm_map['pig_gene'].value_counts() > 1).sum())
print("Mouse genes appearing >1x:", (bm_map['mouse_gene'].value_counts() > 1).sum())
# Both should be 0 for true 1:1
If non-zero, your filter let some non-1:1 entries through.
Check 3 — protein groups with multiple gene names
MaxQuant/DIA-NN sometimes outputs Genes like COL1A1;COL1A2 (protein group has multiple inferred genes). Decide upfront:
- Split and map each → conservative
- Use only the first → simpler, slightly lossy
- Drop multi-gene protein groups → cleanest but loses data
For ECM proteomics with collagens and laminins, multi-gene groups are common. Often the simplest defensible choice is to drop them in the cross-species comparison and report it as a limitation.
Common Pitfalls
"I'm using the wrong species pair"
BioMart asks for "the dataset" (e.g., Sus scrofa) and the homologue species attributes (Mouse Orthologues). The attribute prefix mmusculus_homolog_* is what specifies mouse. Make sure you're not pulling hsapiens_homolog_* (human) by accident — easy mistake when copy-pasting query code.
"Old Ensembl release"
useEnsembl(version = 90) from a tutorial pulls 2017 annotations. Always check which release you are on (the current one is 116 as of October 2026). Use the current release unless you have a specific reproducibility reason to pin an older one — and if you pin, document it.
"Symbol case mismatch in your data"
Mouse genes are conventionally lowercase except first letter (Col1a1). Human and pig conventionally uppercase (COL1A1). BioMart returns them in their native conventions. When joining to your proteomics data, check which case convention MaxQuant used and either match exactly or apply .str.upper() on both sides consistently.
"Some highly conserved proteins missing"
Truly identical sequences across species sometimes lack a "1:1 ortholog" assignment due to annotation gaps. Workaround: cross-reference with OrthoDB or InParanoid if a critical protein is missing from BioMart.
"ortholog_one2one is too strict"
For some studies you may want to include apparent_ortholog_one2one or even ortholog_one2many (collapsing the many-side via summation). Decide based on biological question; default to strict 1:1 unless you have a reason.
Alternatives to BioMart
For most cross-species proteomics, BioMart is the right default. Alternatives worth knowing:
- InParanoid (https://inparanoid.sbc.su.se/): all-vs-all approach, good for less-studied species
- OrthoDB (https://www.orthodb.org/): hierarchical orthology, broader species coverage
- OMA (https://omabrowser.org/): high precision, longer compute
- eggNOG (http://eggnog5.embl.de/): function-aware orthology
Use these when BioMart is missing your species pair or when you need more conservative orthology.
FAQ
Q: Can I use gene-symbol matching for a quick exploration? Yes, for exploration. But once you commit to publishing, switch to BioMart 1:1 — reviewers in cross-species fields increasingly expect it explicitly mentioned in Methods.
Q: What if my data uses Ensembl gene IDs, not gene names?
Even better — Ensembl IDs are stable and unambiguous. Join via ensembl_gene_id ↔ mmusculus_homolog_ensembl_gene and skip the gene name layer entirely.
Q: Does this work for protein IDs (UniProt) instead of gene IDs?
Indirectly. BioMart can provide UniProt IDs as an attribute (uniprotswissprot), and you can join via those. Caveat: some proteins have multiple UniProt accessions per gene (isoforms), so an intermediate gene-level mapping is usually cleaner.
Q: How often should I re-download the ortholog table? Ensembl releases a new version a few times a year. For active long-term projects, re-pull each release and re-validate key orthologs. For a single paper, pin the release at start of analysis and stick with it through publication.
Q: My biomaRt R call is failing with TLS errors.
Common issue with corporate proxies and Ensembl mirror outages. Try httr::set_config(httr::config(ssl_verifypeer = 0L)) for testing, or use a different mirror via mirror = "asia". Production: use the web UI download for one-off, scripted only for reproducibility.
Q: Does this matter for closely related strains within a species? Generally no. Two C57BL/6 mouse substrains use the same annotation — no ortholog mapping needed. Cross-strain or cross-population studies use SNP-based approaches, not orthology.
Closing — The Workflow in One Block
1. Run species-separated proteomics searches (MaxQuant/DIA-NN)
2. Download Ensembl BioMart 1:1 ortholog table (pin release)
3. Filter to ortholog_one2one
4. Sanity-check known orthologs are present
5. Inner-join per-species protein tables via ortholog ID
6. Run cross-species statistics on the joined table
7. Report: BioMart release number + download date + 1:1 filter in Methods
Used this way, cross-species ortholog mapping becomes a transparent, reproducible step instead of a hidden source of errors. The gene-symbol shortcut is fine for exploration; BioMart 1:1 is what survives peer review.
Related posts:
- Reproducing Park et al. 2026 — Cross-Species ECM Proteomics, Three Iterations
- Why You Must NOT Merge Species FASTA Databases in Cross-Species Proteomics
- From DIA-NN Output to Paper Draft: AI-Assisted Proteomics Workflow
- LC-MS/MS Proteomics: Complete Workflow Guide 2026
References:
- Smedley, D. et al. (2015). The BioMart community portal. Nucleic Acids Research, 43, W589-W598.
- Dyer, S. C. et al. (2025). Ensembl 2025. Nucleic Acids Research, 53(D1).
- Sonnhammer, E. L. et al. (2015). InParanoid 8. Nucleic Acids Research, 43, D234-D239.
- Kuzniar, A. et al. (2008). The quest for orthologs. Trends in Genetics, 24, 539-551.
Code and outputs: the scripts and result tables for this post are on GitHub in sbmlab-benchmarks/biomart-pig-mouse.
Related
Why You Must NOT Merge Species FASTA Databases in Cross-Species Proteomics (Shared Peptide Problem)
May 23 · 10 min read
ProteomicsReproducing Park et al. 2026: Three Iterations of a Cross-Species ECM Proteomics Pipeline
May 19 · 13 min read
BioinformaticsBest Bioinformatics Tools in 2026, Installed and Timed: What I Actually Ran This Month
Oct 14 · 7 min read
ProteomicsIs 1% FDR Really 1%? I Hid 14,567 Arabidopsis Proteins in MaxQuant, FragPipe and DIA-NN Searches
Nov 14 · 13 min read