Why BGPT?
logo

The Evidence Layer for Scientific AI

Ask questions across claims linked to experiments, exact results, limitations, and sources.Know what falsifiable science actually supports before you trust the answer.

Press Enter ↡ to search evidence


     Quick Answer



    sciReptor (Imkeller et al., BMC Bioinformatics 2016) was the first antigen-receptor pipeline purpose-built around single-cell data via a relational database linking paired heavy:light chain sequences to flow-cytometry metadata, validated on only three small test datasets (1,152; 1,152; and 384 cells) with no benchmarking against alternatives. Its contribution was structural and durable, but the methodology was never rigorously compared to contemporaneous toolkits.


     Long Answer



    What the paper actually delivers

    sciReptor is a toolkit (Perl/R/Python, MariaDB backend, AGPLv3) that assigns an event ID to every sorted cell, groups heavy, kappa, and lambda sequences by that ID, and links the paired chain calls to user-supplied metadata including flow-cytometry index data β€” capabilities the authors argue existing bulk-oriented tools (IMGT/HighV-QUEST, IGGalaxy, LymAnalyzer, pRESTO, Change-O) lacked . The pipeline chains established components β€” IgBLAST for V/D/J assignment and Change-O-style downstream clonality/hypermutation concepts β€” so the novelty lies in the data model (the cell as a first-class entity in a relational schema), not in new algorithms.

    Critical assessment

    Strengths: The core claim β€” no other ARR tool at the time represented single cells as addressable data structures enabling sequence–phenotype linkage β€” was plausible and well-motivated, and the authors provided a pre-configured virtual machine, manual, test datasets, and public code, which supports reproducibility. The paper explicitly scopes what the software can do rather than overclaiming performance.

    Weaknesses: Validation is demonstrative, not comparative: three datasets totaling ~2,688 sorted cells, one co-authored by the developers themselves (matrix PCR), one Sanger set with a deposited accession range (ENA LN879549–LN879837), and no head-to-head accuracy metrics against any alternative pipeline. Error rates of consensus building, SHM-annotation accuracy, and mispairing frequency are never quantified. The stack (Python 2.7, MariaDB 5.5, Roche/454 dependence) was dated even at publication, and Roche/454's exit from the market severely limits direct applicability today. The claim that no existing tool handles single-cell data is asserted rather than systematically surveyed β€” plausible, but framed by the authors' own platform (citation bias risk). Importantly, the authors acknowledge TCR extension and antigen-binding integration were future work, so reviewers should not treat those as delivered features.

    Confidence and what would change the verdict

    Confidence that the software exists and does what is described: high, given the full text and code repository. Confidence that it was the unique single-cell solution: moderate β€” it rests on the authors' survey of five tools and could be overturned by undocumented contemporaneous workflows. A rigorous external benchmark against Change-O/IgBLAST-native approaches on paired-chain recovery and SHM-calling accuracy, never provided here, is the key missing evidence.

    Author reviews:



    Feedback:    

    Updated: November 19, 2026

     BGPT Paper Review



    Study Novelty

    70%

    First toolkit structurally built around single-cell paired-chain repertoires with relational metadata linkage; the underlying algorithms (IgBLAST, BLAST, MUSCLE, RazerS3) were all pre-existing, so novelty is architectural rather than algorithmic.



    Scientific Quality

    70%

    Honest scope, working code, public test data and VM. However, purely demonstrative validation (~2,688 cells), zero quantitative benchmarking or error-rate reporting, and reliance on soon-obsolete platforms (Roche/454, Python 2.7) temper the quality score.



    Study Generality

    70%

    Applies to human and murine B-cell repertoires across matrix-PCR, conventional PCR, Sanger, and NGS inputs, and was designed to extend to TCR data, but is niche relative to general repertoire tooling and was never generalized broadly in practice.



    Study Usefulness

    80%

    Filled a genuine gap: preserved heavy:light pairing and enabled sequence-to-flow-cytometry linkage, which is essential for interpreting B-cell responses; its database design anticipated later single-cell data integration needs.



    Study Reproducibility

    80%

    Code (GitHub, AGPLv3), step-by-step manual, pre-configured VM, three public test datasets, and an ENA accession for the Sanger set substantially support reproducibility, though exact per-cell validation metrics are not published.



    Explanatory Depth

    50%

    Describes the data model and workflow clearly but offers no quantitative error model, no algorithmic analysis of consensus-calling or SHM-annotation accuracy, and no theoretical treatment of mispairing or cross-contamination rates.


    🎁 Authors: Collect 219 Free Science Tokens (β‰ˆ $21.9 USD)

    Claim My Author Tokens

    Use for 54 days of free BGPT access (4 tokens = 1 day) or trade/sell (β‰ˆ $21.9 USD)

     Top Data Sources ExportMCP



     Analysis Wizard



    Parsing the three published sciReptor test datasets, computing per-cell read-depth and V-J usage distributions, and benchmarking heavy:light pairing recovery against an IgBLAST-based reference workflow to quantify annotation concordance.



     Hypothesis Graveyard



    Existing bulk tools (Change-O, pRESTO) could trivially be adapted for single-cell pairing β€” falsified as a critique because none implemented cell-level identifiers or metadata linkage at the time; the relational design was the missing piece, not a wrapper.


    IgBLAST alone suffices for single-cell annotation β€” falsified because IgBLAST annotates reads per locus but performs no consensus building, barcode demultiplexing, or heavy:light event grouping.

     Science Art


    Paper Review: sciReptor: analysis of single-cell level immunoglobulin repertoires Science Art

     Science Movie



    Make a narrated HD Science movie for this answer ($64 per minute)




     Discussion


    Stay current without chasing every paper.

    Know what changed, what holds up, and what remains uncertain. Every Friday. No ads.


    My BGPT