Skip to content

roadmap.md

Latest commit

 

History

History
195 lines (149 loc) · 5.92 KB

roadmap.md

File metadata and controls

195 lines (149 loc) · 5.92 KB
Raw

Roadmap

v0.1: Assembly Preflight

Goal:

FastaGuard catches FASTA-level assembly problems before expensive assembly QC.

Capabilities:

  • assembly profile
  • plain and gzipped FASTA input
  • streaming parser
  • duplicate IDs
  • duplicate sequences
  • invalid nucleotide/IUPAC symbols
  • empty records
  • core assembly stats
  • N50, N90, L50, L90
  • GC, AT, N, and ambiguity rates
  • high-N scaffolds
  • gap runs
  • tiny contig heuristics
  • length histogram data and HTML plot
  • GC-vs-length plot data and HTML plot
  • PASS / WARN / FAIL verdict
  • JSON, TSV, HTML, and MultiQC-compatible outputs
  • deterministic exit codes

v0.2: Assembly Trust + Pipeline Adoption

Goal:

Make assembly preflight reports easier to trust, route, and consume in real pipelines.

Capabilities:

  • GC, length, and composite assembly outlier findings
  • finding taxonomy, confidence, and follow-up-tool metadata
  • structured routing hints for workflow engines and tool agents
  • richer provenance with command, timestamps, duration, and input size
  • expanded TSV summary rows for outlier counts
  • hardened MultiQC plugin and custom-content support
  • Snakemake and nf-core starter material
  • ready Bioconda recipe metadata for the v0.2.0 update
  • benchmark evidence guidance for adoption decisions

v0.3: Evidence And Assembly Gate

Goal:

Make FastaGuard credible enough for pipeline authors to add as a default assembly gate.

Development scope:

  • public evidence pack from local fixtures and public assemblies
  • Bioconda and BioContainers v0.3 availability documented as live
  • input checksum provenance with provenance.input_sha256
  • clearer machine-readable threshold metadata
  • assembly gate preset with --gate pipeline
  • explicit gate.blocking_findings and gate.advisory_findings for workflow engines and humans
  • clearer blocking vs follow-up recommendations; GC and length outliers remain advisory unless added with --fail-on

v0.4: Preflight Readiness + Compare Mode

Goal:

Make many FASTA files easy to rank, filter, and route before interpretive QC.

Development scope:

  • readiness categories for file, structure, alphabet, index, assembly, submission, and machine readiness
  • index-readiness checks for duplicate first-token IDs and related identifier hazards
  • submission-readiness advisories for FASTA-level issues worth fixing before official validators
  • fastaguard compare as a starter cohort triage layer
  • cross-file metrics table
  • cohort-level outliers
  • sample-to-sample summary
  • combined JSON, TSV, HTML, and MultiQC outputs
  • explicit boundaries: compare mode routes to QUAST, BUSCO, BlobToolKit, CheckM, official validators, annotation, or other downstream tools; it does not replace them

v0.5: Submission Readiness Gate

Goal:

Make assembly FASTA files safer to hand to official validators, annotation, and downstream QC.

Development scope:

  • --gate submission for stricter assembly FASTA preflight
  • --submission-target <generic|ncbi> for target-aware submission advisories
  • stricter identifier and first-token ID safety checks
  • gap-like N run summaries for submission review
  • high ambiguity and tiny-record submission advisories
  • submission readiness fields in JSON, TSV, HTML, and MultiQC outputs
  • compare-mode aggregation of submission readiness across many FASTA files
  • evidence and release notes for --gate submission workflows before official validators
  • clear scope boundaries: FastaGuard does not replace NCBI, ENA, DDBJ, FCS, QUAST, BUSCO, BlobToolKit, CheckM, or annotation validation
  • FastaGuard does not replace NCBI, ENA, DDBJ official validators or guarantee repository acceptance

v0.6: Transcriptome Profile

Potential additions:

  • excessive duplicate transcripts
  • polyA and polyT tails
  • very short transcripts
  • extreme GC outliers
  • isoform-heavy warning heuristics

v0.7: Protein Profile

Potential additions:

  • invalid amino acid symbols
  • internal stop codons
  • terminal stop codons
  • low-complexity regions
  • suspicious nucleotide-looking proteins

v0.8: Reference Panel Profile

Potential additions:

  • stricter ID normalization checks
  • reference naming conventions
  • sequence uniqueness checks
  • panel consistency summaries
  • submission-readiness warnings

Pipeline Integration Maturity

Potential additions:

  • publish and verify package updates for each released contract
  • upstream nf-core module submission
  • official Snakemake wrapper submission
  • Galaxy wrapper
  • upstream MultiQC distribution path
  • BioContainers verification for each published package
  • Homebrew formula

Later Innovation

Avoid external databases in v0.1. Later releases can explore:

  • k-mer sketching with minimizers
  • contamination suspicion without taxonomy
  • optional sourmash, Kraken, or BlobToolKit hooks
  • AI-generated report summaries from local metrics only
  • browser-based interactive contig filtering
  • WASM report viewer

LLM And Tool-Agent Readiness

FastaGuard should prepare for a future where machines talk to QC tools directly. The near-term path is not a chatbot feature; it is a cleaner contract.

Completed foundation:

  • publish schema/fastaguard.schema.json
  • document a stable finding catalog
  • add structured actions[] records to findings
  • add bounded per-record evidence to findings
  • add provenance for profile, thresholds, fail rules, thread count, command, timestamps, duration, and input size
  • add input checksum provenance with provenance.input_sha256
  • add explicit scope fields for what FastaGuard can and cannot conclude
  • add structured routing hints for workflow engines and tool agents
  • add --schema, --finding-catalog, and --explain-finding <id> commands
  • add golden JSON conformance tests

Recommended next sequence:

  • extend evidence tables across submission, transcriptome, protein, reference, and compare modes
  • keep the v0.3 gate contract stable through workflow adoption examples
  • explore an MCP or tool-server interface after the CLI schema is stable