README.md

---
license: cc-by-4.0
pretty_name: MGnify Protein Catalogues
size_categories:
  - unknown
task_categories:
  - other
language:
  - en
tags:
  - biology
  - proteins
  - sequences
  - fasta
  - mgnify
  - metagenomics
---

# MGnify Protein Catalogues

Normalized FASTA shards of MGnify protein cluster catalogues (mgy_clusters and mgy_proteins partitions).

Processed and uploaded by the [MegaData](https://github.com/) post-download pipeline
(internal repo). Original source: <https://www.ebi.ac.uk/metagenomics/>.

> Note: `shard-000000.fasta.zst` per source is a 100-record sample left over from validation passes and is **not part of this dataset**. Only `shard-000001.fasta.zst` and higher contain full-clean data.

## Statistics

| | |
|---|---|
| Source files | 26 |
| Shards | 3,148 |
| Compressed shard bytes | 319.27 GiB (342,815,994,580) |
| Normalized table files | 28 |
| Compressed table bytes | 2.17 TiB (2,380,980,172,257) |

## Layout

```
.
├── tables/<source_slug>.jsonl         # normalized table rows (one JSON object per line)
└── sequences/<source_slug>/shard-NNNNNN.fasta.zst
```

> `metadata/`, `manifests/`, and `_MANIFEST.json` are intentionally not shipped here — the on-disk versions were sample-only. Recompute per-shard record/residue counts downstream by streaming the FASTA records out of the shipped shards.

`<source_slug>` corresponds 1:1 with an upstream source archive; e.g.
`sequence_uniprotkb_uniprot_sprot.fasta.gz`.

## Loading

Stream every shard of one source (replace `<source_slug>` with the directory of
interest under `sequences/`):

```bash
hf download LiteFold/Mgnify --repo-type dataset \
  --include 'sequences/<source_slug>/shard-*.fasta.zst' \
  --local-dir ./mgnify_proteins
zstd -dc ./mgnify_proteins/sequences/<source_slug>/shard-*.fasta.zst | head
```

Programmatic streaming with [`zstandard`](https://pypi.org/project/zstandard/):

```python
from huggingface_hub import snapshot_download
from pathlib import Path
import zstandard as zstd

local = snapshot_download(
    repo_id="LiteFold/Mgnify",
    repo_type="dataset",
    allow_patterns=["sequences/*/shard-*.fasta.zst"],
)

dctx = zstd.ZstdDecompressor()
for shard in sorted(Path(local).rglob("shard-*.fasta.zst")):
    with shard.open("rb") as f, dctx.stream_reader(f) as reader:
        buf = b""
        while chunk := reader.read(1 << 20):
            buf += chunk
            *lines, buf = buf.split(b"\n")
            for line in lines:
                ...  # naive splitter; swap in your FASTA parser
```

## License

CC BY 4.0 (EMBL-EBI MGnify).

## Citation

> Mitchell AL, et al. MGnify: the microbiome analysis resource in 2020. Nucleic Acids Research, 48(D1):D570-D578, 2020.

## Caveats

- `shard-000000.fasta.zst` per source is a 100-record sample left over from validation passes and is **excluded** from this upload. Only `shard-000001.fasta.zst` and onward are shipped.
- Per-source `manifests/*.json` and `metadata/*.records.jsonl` are **not shipped** because their on-disk contents are sample-only (100 records per source). They can be regenerated downstream by streaming the shipped shards (each shard is a concatenation of FASTA records; counting `^>` headers gives per-shard record counts).

## Provenance

Built from the local manifest entry `mgnify_proteins` of `manifests/atlas_download_plan.json`.
Pipeline source: `megadata-post normalize --dataset mgnify_proteins`.