scRepertoire 2: Enhanced and Efficient Toolkit for Single-Cell Immune Profiling

Key results

1. ClaimedCore routines were rewritten in C++ via Rcpp. The authors state that receptor-to-cell pairing in combineTCR/combineBCR now scales linearly rather than quadratically with repertoire size, and that BCR clonotype inference and Levenshtein-based clone clustering also run in linear time.

   Only complexity statements are given: no runtimes, memory figures or comparisons with version 1 or other tools are reported. Other changes: length-threshold filtering before edit-distance clustering, a similar speed-up for amino acid and nucleotide k-mer counting, and export run before plotting when both are requested. · Performance Optimizations; Abstract

2. ClaimedloadContigs auto-detects 10x Genomics, AIRR and BD Rhapsody input, plus MiXCR, TRUST4, WAT3R, Omniscope and Parse Bio Evercode outputs. exportClone writes barcode-level clone tables, including AIRR-compatible ones. Clones from other pipelines can be supplied through a cell metadata column.

   These are described as package features. No import/export test across formats and no round-trip check is reported. Paired clones attach to Seurat or SingleCellExperiment objects for joint analysis with expression data. · Expanded Data Compatibility; Workflow

3. SuggestedOn published erythema migrans data (Jiang et al.), plasma cells showed the strongest clonal expansion. Among each patient's top 20 TCR clones, overlap was observed between lesion and adjacent normal skin, but no BCR clones were shared across tissues.

   Six patients; two samples were excluded (no scAIRR data or no second tissue). 13 lymphocyte types remained after QC. In 192567SKL one clone exceeded 50% of the plasma cell cluster. CD4+ tissue effector memory, cytotoxic CD8+ T and mucosal-associated invariant T (MAIT) cells were also expanded. Among patient 192566's top 20 TCR clones, those shared by blood and lesion did not appear in normal skin. · Results; Figure legends

4. ShownDiversity is estimated with Shannon and inverse Simpson indices, bootstrapped to the smallest repertoire in each comparison, with iNEXT-based interpolation and extrapolation for rarefaction and sample completeness. clonalRarefaction estimates clonal richness by rarefaction and can group samples by condition.

   Each estimate averages 100 bootstrap runs. It was applied per sequencing run to the erythema migrans data (Figure 2D). Differences in diversity between samples are described only in words, with no values or statistical tests. StartracDiversity was reimplemented without custom data dependencies but is not demonstrated. · Results; Clonal Diversity Analysis; Figure legends

5. SuggestedThe new CDR3 sequence summaries (positional amino acid usage, Atchley factors, k-mers, V-J pairing) and Levenshtein-based clone clustering were demonstrated. BCR heavy-chain CDR3s in sample 192567 were rich in serine at positions 6 and 11.

   At those serine-rich positions, mean Atchley factors showed lower molecular size (AF3) and electrostatic charge (AF5). A TRAV usage heatmap was combined with PCA across sequencing runs. TRA chains were clustered at a normalized Levenshtein threshold of 0.85 and shown on a UMAP and as a Fruchterman-Reingold igraph network. All results are descriptive, without statistical testing. · Results; Enhanced Repertoire Summarization; Figure legends