LISA: 2D NMR Lineshape Analysis

📋 On this page

Overview
Workflow at a Glance
Titration Setup
CSP Screen & Isotherm Kd
Spin Systems & ROIs
Global Fit
Results, Bootstrap Errors & Validation
Structure Mapping
Ask AI
Project Files & Export

Overview

LISA (LINESHAPE ANALYSIS) simulates and globally fits the lineshapes of a 2D NMR titration series — HSQC, HMQC, or CPMG-HSQC — to recover molecular interaction (binding) and dynamic (exchange) parameters: dissociation constants (K_d), exchange rates (k_off, k_ex), and the chemical shifts/relaxation of each exchanging state. Its scientific method follows TITAN (Waudby et al., Sci Rep 6:24826, 2016).

Instead of approximating an exchanging resonance as a sum of Lorentzians, LISA treats the experiment as a real virtual spectrometer: it builds the Liouville-space superoperators for the given pulse sequence, adds the population/exchange matrix for a chosen binding model and ligand concentration, and analytically simulates the resulting 2D interferogram. Fitting then optimizes the binding/exchange parameters by global least squares across all residues and titration points at once — not one peak at a time.

💡 Where to find it: Extensions menu → LISA: 2D NMR Lineshape Analysis (accelerator LS). The dialog is non-modal — you can keep selecting peaks in the main spectrum view while it stays open.

LISA complements CAFE: CAFE's CSP/K_d fitting works directly from peak positions and is fast, but assumes fast exchange and treats each residue independently. LISA fits the full 2D lineshape — informative in slow, intermediate, and fast exchange, and recovers k_off as well as K_d from one global fit across every residue.

Workflow at a Glance

The dialog has five tabs, used roughly left to right:

Tab	Purpose
Titration Setup	Pick the titration series' spectra, enter concentrations and acquisition parameters
CSP Screen	Fast triage — which residues actually respond to the ligand — and quick isotherm K_d fits
Spin Systems & ROIs	Define the fitted region (ROI) per residue, across every titration point
Fit	Edit initial parameter guesses/bounds, run the global least-squares fit
Results	Observed/fitted/residual contours, 1D projections, validation, bootstrap errors, exports

Titration Setup

Pick titration spectra... selects every point of the series in titration order (apo first) — order matters, since the ligand/protein concentrations entered next are matched positionally. Acquisition parameters (¹⁵N/¹H spectrometer frequencies, t1 points) auto-fill from the picked spectra; verify them, and enter the exponential line-broadening (LB Hz) applied during processing yourself — that value isn't recoverable from the spectrum file itself.

⚠ Units: concentrations are entered in M (e.g. 1.0e-6 for 1 µM), not µM directly — LISA converts internally, but the entry field itself expects molar.

CSP Screen & Isotherm K_d

Before investing in a full lineshape fit of every residue, the CSP Screen tab triages which spin systems are actually perturbed by the ligand: it flags any system whose combined chemical-shift perturbation exceeds an iteratively-refined mean + N·σ noise threshold (default σ multiplier 3.0), then lets you Promote selected systems on to the Fit tab.

The same tab fits a quick CSP isotherm K_d directly from chemical-shift perturbations — a hyperbolic (1:1, with protein-depletion correction) or Hill model, whichever suits the shape of the data — far cheaper than a full lineshape fit, and a useful sanity check against the Fit tab's result once both are available. This mode requires shift data: capture it on the Spin Systems & ROIs tab first (see below).

Spin Systems & ROIs

LISA fits regions of interest (intensity boxes) around each peak, not whole spectra — the same idea as POKY's A_MAP. Select a peak (or several) in the main spectrum view, then New ROI from selected peaks creates a tile in the ROI grid and a row in the spin-systems table for each. For a project with many ROIs, the tile grid is windowed/paginated (step buttons move through residues and titration points) rather than rendering every tile at once.

Capture shift from selected peak (one peak) or Capture all points (by assignment) (every titration point at once, matched by residue/atom assignment) records each spin system's observed peak position per point — the input the CSP Screen tab needs, and a useful cross-check on the Fit tab's own recovered shifts.

💡 Most common cause of a bad fit: an ROI tile that isn't actually centered on the peak at every titration point. Check the grid across the whole series before running the fit, not just at the apo point.

Global Fit

The parameter table lists every fitted quantity — global (K_d, k_off, shared across all residues) and local (per-residue free/bound shifts and linewidths) — with its current value and, for K_d/k_off, an editable physical bounds range. Single-click a row's Value to edit the initial guess, its Fixed cell to hold a parameter constant (e.g. a K_d already known from another method), or its Bounds cell to constrain K_d/k_off to a physically sensible range.

💡 Set K_d/k_off bounds: the global fit is a nonlinear optimization and can, for some parameter combinations, converge on an unphysical result (K_d near zero or near infinity) if left completely unbounded. Entering a reasonable physical range — e.g. K_d between 1 nM and 1 mM for a typical protein–ligand interaction — guards against this and costs nothing if your fit was already landing in that range anyway.

Run fit executes the paper's staged procedure: Stage 1 fits free-state shifts/linewidths from the apo point alone (no binding/exchange information available yet), Stage 2 then fits everything — linewidths at both states plus the global K_d/k_off — across the complete series. Progress is reported live; Stop cancels a running fit cleanly.

Results, Bootstrap Errors & Validation

The Results tab shows, per spin system and titration point: observed vs. fitted 2D contours with a residual map, 1D projections (both dimensions' marginals), and 1D cross-sections through the ROI center. Validate fit... runs residual diagnostics — reduced χ², R², and a structured-residual check — before you trust any recovered K_d/k_off at face value.

Back on the Fit tab, Bootstrap errors (after a successful fit) resamples the converged fit's residuals in configurable tile blocks to report a standard error alongside every fitted parameter — the paper's validated approach to parameter uncertainty for spatially-correlated 2D NMR residuals.

Structure Mapping

Map to Structure... (Results tab) takes a loaded PDB file and writes a PyMOL .pml script coloring residues by their fitted chemical-shift-perturbation magnitude — the same idea as CAFE's structure export, driven here by the lineshape fit's own recovered per-residue values rather than a raw CSP.

Ask AI

Ask AI... (Fit tab) sends a question you write yourself, prefixed with LISA domain context (including the current fit's own results and diagnostics), to an LLM for interpretation help — e.g. "is this k_off/Δω ratio consistent with fast exchange?" Nothing is sent automatically; the request goes out only when you submit a question, and the payload is the same fit-report text shown in the dialog first. Use it for interpretation, not as a numerical authority.

Project Files & Export

Save project.../Open project... (Titration Setup tab) round-trips the whole session — titration series, acquisition parameters, every spin system's ROIs and captured shifts, and the current parameter/fit state — as a single JSON project file. Spectra are referenced by name, so open the same spectra in POKY before opening a saved project.

Export params... and Export results... (Fit tab) write the parameter table and per-point fit results to CSV; Export figures... (Results tab) saves the plot panels — all ready for a spreadsheet or a manuscript figure.