Skip to content

Peak picking & fitting

Peak picking

Find peaks in 1D to 3D spectra with a choice of classical and neural pickers, then correct them by hand.

Automated picking is a recall-versus-precision trade, and the right point on that curve depends on what happens next. Assignment wants everything real, including the weak peaks, and tolerates a few artefacts you will reject later. Quantitative fitting wants only peaks whose intensities you intend to trust.

So we do not pick with one algorithm. A classical threshold-and-cluster picker, a neural picker trained on real spectra, and our own clustering picker are all available against the same spectrum, and their outputs are directly comparable because they land in the same peak-list format. In practice the useful move is to run two and look at where they disagree.

Whatever picks, you correct. The viewer supports adding, removing and dragging peaks against live contours, with the peak list updating underneath, because no picker gets a crowded amide region entirely right.

What it needs

  • Processed spectrum
  • Optionally, an existing peak list to refine

What you get

  • Peak list in Sparky, NMRPipe, NEF or CSV format

What does the work

  • PICA

    Built in-house

    Our picker and cluster deconvolution, 1D to 3D

  • iPick

    Open source, redistributed under its licenceBSD-2-Clause

    Classical threshold picking

  • DEEP Picker

    Open source, redistributed under its licenceGPL-3.0

    Convolutional peak detection

Related in Peak picking & fitting