Spectrum-Level Prompts

Example 1: Score one selected PSM

User prompt

Take the highest-confidence PSM for PEPTIDESEQ, load the matching spectrum, predict fragments with MS2PIP, and score it.

What the MCP should do

  • inspect the table first to identify the target PSM
  • load or retrieve the observed spectrum
  • call score_spectrum
  • optionally call summarize_score_metrics

Expected output shape

  • one SpectrumMetricResult
  • grouped metrics such as match_quality, predicted_match_evidence, mass_error, and acquisition

Example 2: Render a mirror plot

User prompt

Render a mirror plot for scan 12345 from run sample_A using the candidate sequence from this table.

What the MCP should do

  • identify the matching candidate row
  • score the spectrum or build a payload from observed and predicted spectra
  • call render_spectrum_comparison

Expected output shape

  • standalone HTML artifact
  • path to the rendered file

Example 3: Annotate observed ions only

User prompt

Annotate the observed spectrum for this phosphopeptide using all ion types, including neutral losses, without showing the predicted spectrum.

What the MCP should do

  • call annotate_observed_spectrum or render_annotated_observed_spectrum
  • use ion_types="all"
  • use neutral_losses=true

Expected output shape

  • observed-spectrum annotation summary
  • optionally an annotated observed-only HTML view

Example 4: Compare several selected PSMs

User prompt

Compare the spectral angle, matched-ion fraction, explained intensity, and log2 xcorr for these five selected PSMs.

What the MCP should do

  • score each selected target
  • summarize the returned metric groups across the subset

Expected output shape

  • one short comparison table or structured list of per-PSM metrics
  • no need to render HTML unless requested