PECAN/CheSPI — POKY Suite Manual

📋 On this page

Overview
PECAN tab
CheSPI tab
CheSPI input options
Reading CheSPI's results
References

Overview

PECAN/CheSPI (command n6) is a two-tab dialog that predicts protein secondary structure directly from assigned chemical shifts — no structure calculation or NOE data required. The PECAN tab runs POKY's own density-based classifier (helix/strand probability per residue); the CheSPI tab runs a POKY-adapted build of Nielsen & Mulder's CheSPI, which additionally estimates helix/sheet/turn/coil populations (useful for partially-folded or intrinsically disordered regions) and full 3-state and 8-state DSSP class probabilities.

Access it from Utilities → PECAN/CheSPI, or type its command code n6.

💡 Reference your shifts first. Both methods are sensitive to chemical shift referencing errors. Consider running pyLACS (lv) before either tab if you haven't already re-referenced your data.

PECAN tab

Run predicts secondary structure from your current session's chemical shifts (CA, CB, HA, C). You'll be asked which condition to use and whether to include unused resonances, then for a directory to write results into. BMRB entry runs the same prediction against a local NMR-STAR file or a published BMRB entry number instead of the current session.

Results are written as an NMR-STAR-style pecan_results.txt (opened automatically in Notepad) with per-residue extended/helix probabilities, and can optionally be plotted with POKY's built-in NDP plot tool.

CheSPI tab

Run predicts secondary structure from the current session's chemical shifts, the same way as PECAN's Run button. BMRB entry lets you instead point CheSPI at a local NMR-STAR file or download a published BMRB entry directly.

Once a run finishes, the output box shows a compact "short summary" line (sequence, per-residue confidence digit, and 8-class DSSP string) followed by a full per-residue table: amino acid, best-matching 8-state DSSP class, its confidence, and the helix/sheet/turn/coil population percentages.

Save results... writes the summary and table to a text file and opens it in Notepad.
Export PyMOL coloring... saves a .pml script that colors each residue by its CheSPI-derived color (from the CheZOD Z-score and principal components) and copies a run <path> command to your clipboard — load your structure in PyMOL, then paste that command into PyMOL's command line.

ℹ️ CheSPI needs no external install. It runs as a self-contained script bundled with POKY (adapted for NMR-STAR 3.1 input by the POKY team from the original CheSPI, which is itself built on the embedded random-coil shift predictor POTENCI) — no separate download, server round-trip, or account is needed.

CheSPI input options

Source	What's used	When to prefer it
Run (current session)	Sequence + shifts from your open project	Everyday use while assigning
BMRB entry → Local file	An existing NMR-STAR file as-is	You already have a full deposition (e.g. from I-PINE) with real sample conditions
BMRB entry → BMRB number	Downloaded directly from bmrb.io	Checking a published entry, or comparing against your own data

⚠️ Sample conditions. CheSPI uses pH, temperature, and ionic strength to predict random-coil reference shifts (via POTENCI) - getting these wrong biases the result, since it's the deviation from the predicted random-coil shift that CheSPI actually interprets. When running from the current session, POKY asks for these three values before each run (POKY doesn't track them as session metadata, so they can't be pulled automatically) - use your sample's real values, not the defaults, if you know them. BMRB entries and local NMR-STAR files that already specify sample conditions are used as-is instead.

⚠️ Chemical shift referencing. Like PECAN, CheSPI assumes your shifts are referenced to DSS (the standard for solution biomolecular NMR) and includes its own automatic re-referencing check against the predicted random-coil shifts. If your data isn't DSS-referenced, consider correcting it first with pyLACS (lv).

Reading CheSPI's results

The 8-state DSSP classes are: H alpha-helix, G 3-10-helix, I pi-helix, E sheet, - none (extended/disordered), T turn, S bend, B bridge. The population percentages (helix/sheet/turn/coil) are the more informative view for partially structured or disordered regions, since a residue rarely sits at 100% in any one class — CheSPI is specifically designed to resolve small amounts of residual structure that a single best-class call would hide.

References

Nielsen, J.T. and Mulder, F.A.A. CheSPI: chemical shift secondary structure population inference. J. Biomol. NMR 75, 273–291 (2021). doi:10.1007/s10858-021-00374-w

CheSPI's random-coil shift predictions are computed via POTENCI (embedded in the same tool), and its original implementation is available at github.com/protein-nmr/CheSPI. If CheSPI results contribute to your publication, please cite the paper above.