Skip to content

Performance Guidance

This section provides recommendations for optimising performance across VariantValidator SHAIP endpoints.

Performance is primarily influenced by:

  • Transcript selection strategy
  • Use of liftover
  • Dataset scope (all vs targeted)
  • Inclusion of auxiliary data (e.g. exon structures)

variantformatter-v2

Recommended configuration for optimal performance:

  • Use liftover = false
  • Select transcript_model = refseq or ensembl
  • Use:
  • mane_select (fastest; MANE Select only), or
  • mane (MANE Select + MANE Plus Clinical, slightly broader)
  • Alternatively specify explicit transcript IDs
  • (see transcript_selection.md)
  • Avoid all or raw unless required

Notes

  • This is the fastest endpoint for genomic inputs
  • Designed for high-throughput processing
  • Performance decreases when enumerating large transcript sets

variantvalidator / variantvalidator-ensembl

Recommended configuration:

  • Prefer specific transcript IDs where possible
  • Otherwise use:
  • mane_select (MANE Select only; most efficient curated set)
  • mane (MANE Select + MANE Plus Clinical; broader but still controlled)
  • Alternatively specify explicit transcript IDs
  • (see transcript_selection.md)
  • Avoid all and raw unless required
  • Submit single variants or small batches

Notes

  • These endpoints perform full HGVS validation and are more computationally intensive
  • Suitable for complex HGVS inputs and error-tolerant parsing
  • Transcript restriction is the primary determinant of performance

gene2transcripts-v2

Recommended configuration:

  • Use:
  • mane_select (small, curated transcript set), or
  • mane (expanded curated set including MANE Plus Clinical)
  • Alternatively specify explicit transcript IDs
  • (see transcript_selection.md)
  • Avoid all and raw unless full enumeration is required
  • Set:
"show_exon_info": false

when exon structure data is not required

Notes

  • Including exon data significantly increases response size and processing time
  • Using all can produce very large outputs
  • Best used as a filtering step prior to validation or formatting

Summary

Fastest configurations across endpoints:

  • Restrict transcripts using mane_select, mane, or explicit IDs
  • Avoid all and raw
  • (see transcript_selection.md)
  • Disable liftover where not needed
  • Exclude exon information unless required

These practices minimise computational overhead and maximise throughput in pipeline and batch-processing environments.