Guide for NMR Spectroscopists

How to use synth-pdb to generate structures with realistic synthetic NMR observables for training datasets, validation pipelines, and spectral simulation.

Quick-Start: Generating a Structure with All NMR Data

synth-pdb \
  --sequence MGSSHHHHHHSSGLVPRGSH \
  --minimize \
  --gen-nef \
  --gen-relax \
  --gen-shifts --shift-predictor shiftx2 \
  --gen-couplings \
  --output-rdcs rdcs.csv \
  --output peptide.pdb

This produces: - peptide.pdb — coordinates - peptide_nef.nef — NOE restraints in NEF format - peptide_relax.nef — R₁, R₂, NOE relaxation data in NEF - peptide_shifts.nef — ¹H, ¹⁵N, ¹³Cα, ¹³Cβ, ¹³C chemical shifts in NEF - peptide_couplings.csv — ³J(HN,HA) couplings for each residue - rdcs.csv — backbone ¹D_NH RDCs for each residue

Controlling Chemical Shift Prediction: `--shift-predictor`

When --gen-shifts is active, you choose which predictor backend synth-nmr uses.

Flag	Backend	Best For
`--shift-predictor shiftx2`	SHIFTX2 (Han et al., 2011)	Best accuracy; needs binary
`--shift-predictor empirical`	SPARTA+ / empirical tables	Reproducible CI, no external binary

Default is shiftx2, which automatically falls back to the empirical method if the SHIFTX2 binary is not installed.

# Use empirical method explicitly (e.g., in Docker CI where SHIFTX2 not installed)
synth-pdb --sequence ACDEFGHIKLM --gen-shifts --shift-predictor empirical \
  --output peptide.pdb

Accuracy comparison (RMSD from BMRB experimental shifts):

Nucleus	SHIFTX2	SPARTA+
¹H	0.04 ppm	0.05 ppm
¹³C	0.44 ppm	0.55 ppm
¹⁵N	1.17 ppm	2.06 ppm

Sources: Han et al. (2011), Shen & Bax (2010)

Generating Synthetic RDC Data: `--output-rdcs`

RDCs are backbone \({}^1D_\text{NH}\) values (Hz), computed from the N–H bond vector orientation relative to an alignment tensor.

synth-pdb \
  --sequence ACKNILQ \
  --minimize \
  --output-rdcs rdcs.csv \
  --rdc-da 12.0 \
  --rdc-r 0.15 \
  --output peptide.pdb

Note: --minimize is recommended with --output-rdcs to ensure the structure has backbone amide H atoms added by OpenMM.

Alignment Tensor Parameters

Flag	Default	Meaning
`--rdc-da`	`10.0`	Axial component \(D_a\) in Hz (typical: 5–25 Hz)
`--rdc-r`	`0.1`	Rhombicity \(R\), \(0 \leq R \leq 2/3\)

The RDC formula (Tjandra & Bax, 1997, Science 278:1111):

\[D(\theta, \phi) = D_a\left[(3\cos^2\theta - 1) + \frac{3}{2}R\sin^2\theta\cos 2\phi\right]\]

RDC Output CSV Format

res_id,residue,RDC_NH_Hz
1,A,18.4231
2,C,7.1203
3,K,-3.8820
4,N,12.4455
...

Proline residues are automatically excluded (no backbone amide H).

Physical Range

For \(D_a = 10\) Hz, \(R = 0.1\): values span approximately \(-11.5\) to \(+20\) Hz. Values outside the range \([D_a(-1-1.5R),\; 2D_a]\) indicate an error.

Interpreting the PDB REMARK Header

Every synth-pdb output PDB file embeds a REMARK 3 block recording the exact command-line invocation:

REMARK 3  GENERATION PARAMETERS:
REMARK 3  Command:
REMARK 3    synth-pdb --sequence ACKNILQ --minimize --output-rdcs rdcs.csv
REMARK 3    --rdc-da 12.0 --rdc-r 0.15 --gen-shifts --output peptide.pdb

This satisfies FAIR data principles (Findable, Accessible, Interoperable, Reusable): any researcher can exactly reproduce the generation from the REMARK block alone.