{"success":true,"data":{"author":"Briney Lab, Scripps Research; source data Jaffe et al. 2022 (10x Genomics)","bucketAccessType":"public","bucketName":"brineylab/jaffe-paired","bucketPrefix":"","category":"Biology","cloudProvider":"huggingface","createdAt":"2026-11-22T01:09:20.836Z","description":"Natively paired human antibody heavy and light chains from the single-cell B-cell receptor (BCR) sequencing study of Jaffe et al. 2022, Functional antibodies exhibit light chain coherence (Nature), rebuilt by the Briney Lab at Scripps Research as a machine-learning-ready set. The source is the Cell Ranger V(D)J contigs from 94 samples of 4 donors in figshare article 19617633 (DOI 10.6084/m9.figshare.19617633.v3). Every contig was annotated with abstar 0.9.0 against the human germline; droplets were kept only with exactly one productive heavy (IGH) and one productive light (IGK or IGL) chain; about 1.39 million pairs were clustered with MMseqs2 at 85% identity on the joined variable domains and split by cluster 96/2/2 into train, eval and test, with train reduced to 95% cluster representatives. Each Parquet row carries the pair ID, donor, sorted B-cell class (naive, unswitched, switched, plasmablast), cluster IDs and around 340 AIRR fields per chain such as v_call, cdr3_aa and sequence_vdj_aa. Use it to train paired antibody language models or study light chain pairing; load with polars or pandas. License MIT, as stated by the figshare source. Cite Jaffe et al. 2022.","displayTitle":"Jaffe 2022 Paired Human Antibody Heavy-Light Chains from Single-Cell 10x Sequencing, Clustered Splits (Briney Lab)","huggingFaceUrl":"https://huggingface.co/datasets/brineylab/jaffe-paired","id":"0485929c-666a-5bae-b68d-de306b143cf2","license":"MIT","licenseEvidenceUrl":"https://figshare.com/articles/dataset/19617633","objectCountEstimate":5,"probeStatus":"ok","referenceUrl":"https://huggingface.co/datasets/brineylab/jaffe-paired","region":"us","schemaCard":{"accessHints":[{"language":"python-datasets","snippet":"from datasets import load_dataset\nds = load_dataset(\"brineylab/jaffe-paired\")"},{"language":"python-hub","snippet":"from huggingface_hub import snapshot_download\nsnapshot_download(repo_id=\"brineylab/jaffe-paired\", repo_type=\"dataset\")"}],"caveats":["HuggingFace metadata: author=brineylab, license:mit, size_categories:1M<n<10M, format:parquet, downloads=87"],"format":"mixed","generatedAt":"2026-11-22T01:10:21.365Z","layout":{"example":"eval.parquet","pattern":"{repoId}/{filename}"},"mountPath":"/mnt/archive","sampleHeaders":[{"bytes":"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","encoding":"hex","key":"eval.parquet","truncatedAt":512}],"sampleObjects":["eval.parquet","test.parquet","train.parquet"],"sidecars":[{"bytes":"---\nlicense: mit\npretty_name: jaffe-paired\n---\n# jaffe-paired\n\nHub: `brineylab/jaffe-paired`, tag `v1.0.0`\n\n## Files\n\n`train.parquet`, `eval.parquet`, `test.parquet`: one row per pair, a heavy and a light chain from one droplet. Columns:\n\n- `pair_id`: `<barcode>_<sample>` (`AAACCTGAGACACGAC-1_1279049`).\n- `sample`, `donor` (1 to 4), `flow_class` (`naive`, `unswitched`, `switched`, `plasmablast`, `unswitched + naive`, `switched + naive`): the sample's row of the source's `test_data.csv`.\n- `cluster_85`: the `pair_id` of the pair's 85% cluster representative. In eval and test every row is its own representative.\n- `cluster_95` (train only): the `pair_id` of its 95% representative; every train row is its own.\n- Every field abstar 0.9.0 writes, once per chain: the heavy chain's with `:0` appended (`v_call:0`, `cdr3_aa:0`, `sequence_vdj_aa:0`), the light chain's with `:1`. `sequence_vdj_aa` is the variable domain, FWR1 through FWR4.\n\n## Reading it\n\n```python\nfrom huggingface_hub import snapshot_download\nimport polars as pl\npath = snapshot_download(\"brineylab/jaffe-paired\", repo_type=\"dataset\", revision=\"v1.0.0\")\ntrain = pl.read_parquet(f\"{path}/train.parquet\", columns=[\"pair_id\", \"sequence_vdj_aa:0\", \"sequence_vdj_aa:1\"])\n```\n\nName the columns you need: each file holds some 340.\n\n## Recipe\n\n**Source:** Jaffe et al. 2022, figshare article 19617633 version 3 (DOI `10.6084/m9.figshare.19617633.v3`, MIT), pulled 2026-11-06: `test_data.tar.gz` (file 34846953, MD5 `4d87c4da2d407ba95180be2cb1c91044`), of which each sample's Cell Ranger `vdj_b/filtered_contig.fasta` (cell-associated contigs, productive, high-confidence, and full-length by Cell Ranger), and `test_data.csv` (file 34846938, MD5 `3ccae335a8a03b5a771dc0eae97f7af6`), the 94 samples' donors and flow classes. The recipe refuses a download whose MD5 differs.\n\n**Steps**, one job in `datascience@1.2.0`:\n\n1. Annotate every contig with abstar 0.9.0 (`germline_database = \"human\"`). 0.9.0 matters: it writes `sequence_vdj_aa`, the variable domain, which the clustering uses; in 0.9.0 `sequence_aa` is the translated input, as AIRR 2.0 defines it.\n2. Pair within each sample: a droplet's barcode is the contig id up to its first underscore. Keep a droplet only with exactly one heavy chain (`IGH`) and one light (`IGK` or `IGL`), both productive by abstar. The build prints what it dropped, by shape (`dropped 1H2L: <n>`) and for productivity (`dropped unproductive: <n>`); the record carries those lines.\n3. Cluster every pair with `mmseqs easy-cluster --min-seq-id 0.85 -c 0.8 --cov-mode 0` (`identity`, `coverage`; 80% of both sequences) on the heavy chain's `sequence_vdj_aa` joined to the light chain's. One `easy-cluster` run over every pair: `datascience@1.2.0`'s mmseqs (master, commit 2a3dafc) carries the fix to the prefilter overflow (MMseqs2 #1092) that kills release 18 and abutils' bundled build on about 1.1M of these sequences.\n4. Split by cluster `[0.96, 0.02, 0.02]`: the representatives, sorted, are shuffled with `random.Random(seed = 1)`, and whole clusters go to train until it holds 96% of the pairs, then to eval until 2%, then to test. No 85% cluster spans two splits. Eval and test keep the 85% representatives.\n5. Re-cluster every train pair at 95% (`train_identity`), the same way; train keeps the 95% representatives.\n\n`params.toml` also holds `limit = 0`, the number of samples for a developer's dry run; 0 is all.\n\n**The test** checks the three files and their columns, each split's rows against a floor, no `pair_id` or `cluster_85` in two splits, both chains productive, `locus:0` `IGH` and `locus:1` `IGK` or `IGL`, and each `sequence_vdj_aa` free of gaps and stops and ending with its chain's `fwr4_aa`.\n\n## Compute\n\nOne job, CPU only: 32 CPUs and 200G asked, on coreweave-rtx6000 `hpc-low` (Xeon Platinum 8562Y+). From nothing it takes about 4 h 20 min: the download (12 GB) 6 min; abstar and the pairing, sample by sample, 2 h 10 min; the 85% clustering of all 1,389,546 pairs 7 min; the 95% clustering of train's 1.33M 1 h 50 min. Peak memory 81 GB, in the annotation. Disk: 1.5 GB of f","encoding":"utf8","key":"README.md","name":"README.md","size":4961,"truncated":true}],"sourceObjectKey":"eval.parquet","starterCell":"from datasets import load_dataset\nds = load_dataset(\"brineylab/jaffe-paired\")\nprint(ds)","starterSamplePath":"eval.parquet","totalObjects":5},"summary":"Natively paired human antibody heavy and light chains from the single-cell B-cell receptor (BCR) sequencing study of Jaffe et al.","title":"Jaffe 2022 Paired Human Antibody Heavy-Light Chains from Single-Cell 10x Sequencing, Clustered Splits (Briney Lab)","updatedAt":"2026-11-22T03:37:38.394Z","visibility":"public","websiteUrl":"https://huggingface.co/datasets/brineylab/jaffe-paired"}}