py-scoper

Pure-Python port of the R/CRAN package scoper — Spectral Clustering for clOne Partitioning — for identifying B-cell clones from Adaptive Immune Receptor Repertoire sequencing (AIRR-Seq) data.

scoper is part of the Immcantation framework, developed by the Kleinstein Lab at Yale University. py-scoper re-implements its three clonal-partitioning methods faithfully in numpy / scipy / pandas / matplotlib — no rpy2, no R runtime required.

import pyscoper as sc

db = sc.load_example_db()                      # scoper's ExampleDb (2000 BCRs)

ident = sc.identicalClones(db)                 # clones by identical junction
hier  = sc.hierarchicalClones(db, threshold=0.15)
spect = sc.spectralClones(db, method="novj")   # scoper's signature method

print(spect["clone_id"].nunique(), "clones")

Each function returns the input AIRR data frame with an integer clone_id column.

Installation

pip install pyscoper

Optional extras pyalakazam / pyshazam supply amino-acid translation, germline-consensus and locus-inference helpers (pip install pyscoper[full]). py-scoper ships internal fallbacks, so they are not strictly required.

Methods

With summarize_clones=True a ScoperClones object is returned, carrying the per-group summary table, the inter- vs intra-clone distance table, the effective clonal-distance threshold, and a plotCloneSummary plot.

res = sc.spectralClones(db, method="novj", summarize_clones=True)
res.summary()              # per V/J/length group statistics
res.eff_threshold          # effective distance threshold
sc.plotCloneSummary(res)   # clonal-distance distribution plot

R parity

py-scoper is validated against scoper 1.5.0 on its bundled ExampleDb (2000 heavy-chain BCR sequences). Agreement is measured by the Adjusted Rand Index (ARI) between the R and Python clonal partitions:

The spectral method uses k-means with random initialisation, so an exact partition is not expected; the parity tests require ARI > 0.95 and the total clone count to be within a few percent of scoper's.

Run the parity suite (needs an R install with scoper):

pytest tests/ -q

Public API

identicalClones, hierarchicalClones, spectralClones, defineClonesScoper, ScoperClones, plotCloneSummary, calculate_inter_vs_intra, effective_threshold, fast_dist, pairwise_dist_dna, pairwise_dist_aa, pairwise_mut_matrix, count_invalid_bases, group_genes, pass_to_clustering, krnl_mtx_generator, laplacian_mtx, make_affinity, likelihoods, infer, find_gap_smooth, range_a_to_b, gaussian_kde_r, bw_nrd0, load_example_db.

License

AGPL-3, matching the original scoper package. See LICENSE. Credit to the Immcantation framework and the Kleinstein Lab (Yale University).

References

• Nouri N, Kleinstein SH (2018). A spectral clustering-based method for identifying clones from high-throughput B cell repertoire sequencing data. Bioinformatics, i341–i349.
• Nouri N, Kleinstein SH (2020). Somatic hypermutation analysis for improved identification of B cell clonal families from next-generation sequencing data. PLOS Computational Biology, 16(6), e1007977.
• Gupta NT, et al. (2017). Hierarchical clustering can identify B cell clones with high confidence in Ig repertoire sequencing data. The Journal of Immunology, 2489–2499.