Initial release

[kevindetry-milaboratories]

Greptile Summary

Initial release of an end-to-end generation-probability block.

• Adds a Python/OLGA scoring service that reads clonotypes from Parquet and computes raw and −log10 generation probabilities in parallel.
• Adds Tengo workflows that discover CDR3 and chain columns, invoke the scoring service, and export scored PFrames.
• Adds model outputs and Vue pages for dataset/species selection, tabular results, progress reporting, skipped-chain warnings, and probability distributions.
• Adds package metadata, dependencies, release notes, and user-facing documentation.
• Important touched terms:
  • Pgen (generation probability): Probability that V(D)J recombination generates a given CDR3; introduced as the primary computed value and exported column.
  • −log10(Pgen): Log-transformed rarity score where larger values indicate rarer CDR3s; introduced as a companion output.
  • CDR3: Receptor sequence region scored by OLGA; workflow discovery now supports VDJ clonotype and variant sequence namespaces.
  • Chain: Immunoglobulin or T-cell receptor locus associated with a CDR3; used to select the appropriate recombination model.
  • Species: Organism whose recombination model is used; introduced as a required human-or-mouse setting.
  • Alphabet: Amino-acid or nucleotide representation of a CDR3; inferred from dataset metadata and passed to the scorer.
  • PFrame: Platforma columnar frame carrying scored results; introduced for table, graph, and export outputs.
  • OLGA: Library that evaluates generation probability using learned recombination models; added as the scoring engine.

Confidence Score: 5/5

The pull request appears safe to merge with no concrete blocking or independently actionable non-blocking defects identified.

The implemented model, workflow, scorer, and UI form a consistent data path from selected repertoire columns through OLGA scoring to tabular, graphical, and exported results.

Important Files Changed

Filename	Overview
software/src/main.py	Implements validated Parquet input handling, model selection, parallel OLGA scoring, transformed scores, and progress reporting.
workflow/src/main.tpl.tengo	Resolves dataset key and alphabet metadata, invokes the scoring workflow, and publishes cached outputs.
workflow/src/score.tpl.tengo	Discovers scoreable CDR3/chain units for bulk and single-cell datasets and constructs the scoring input.
workflow/src/pgen-process.tpl.tengo	Executes the Python scorer with bounded resources and converts its result into display and export PFrames.
model/src/index.ts	Defines block state, input discovery, workflow arguments, tables, graph frames, progress, and navigation.
ui/src/pages/MainPage.vue	Adds dataset/species settings, scored-result table, and unsupported-chain feedback.
ui/src/pages/DistributionPage.vue	Adds a histogram view over generated raw Pgen columns.

Sequence Diagram

sequenceDiagram
    participant U as User
    participant UI as Vue UI
    participant M as Block Model
    participant W as Tengo Workflow
    participant P as Python/OLGA
    U->>UI: Select dataset and species
    UI->>M: Persist block arguments
    M->>W: Run with anchor and species
    W->>W: Discover CDR3 and chain columns
    W->>P: Score Parquet input
    P-->>W: Result Parquet and progress
    W-->>M: PFrame outputs
    M-->>UI: Table, histogram, warnings

Loading

_{Reviews (1): Last reviewed commit: "Initial release" | Re-trigger Greptile}

[gemini-code-assist]

Caution

The consumer version of Gemini Code Assist on GitHub has been sunset. All code review activity has officially ceased.