Skip to content

nerd tempgrad_fit

nerd tempgrad_fit performs temperature-gradient analyses (Arrhenius or two-state melt) using the pluggable engines under pipeline/plugins/tempgrad. It can consume NMR fits (data_source: nmr) or probe timecourse fits (data_source: probe_tc) and stores results in tempgrad_fit_runs/tempgrad_fit_params.

nerd tempgrad_fit --config PATH/TO/config.yaml --db PATH/TO/nerd.sqlite

If --db is omitted, NERD falls back to <run.output_dir>/nerd.sqlite.

Configuration Layout

run:
  label: tempgrad_demo
  output_dir: examples

tempgrad_fit:
  mode: arrhenius                 # or two_state_melt
  data_source: probe_tc           # nmr or probe_tc
  overwrite: true

  filters:
    construct: 4U_wt
    base: A
    melt_threshold_c: 60
    fit_kind: round3_constrained

  engine_options:
    temperature_unit: c
    weighted: true

Key Fields

Field Description
mode Selects the engine (arrhenius, two_state_melt).
data_source nmr (uses nmr_fit_runs) or probe_tc (uses probe_tc_fit_params).
filters Constraints on construct, buffer, base, nt, melt threshold, etc.
engine_options Passed to the engine; e.g., weighted, temperature_unit, melt_threshold_c.
overwrite Removes existing rows in tempgrad_fit_runs/tempgrad_fit_params before inserting new ones.

For probe_tc, grouping happens automatically: reactions are grouped by construct, buffer, probe, concentration, and RT protocol; temperatures are filtered using melt_threshold_c before fitting each nucleotide.

Supported Data Sources

  1. NMR Arrhenius (data_source: nmr)
    Pulls completed fits from nmr_fit_runs, using k_value/k_error to produce log-linear fits.

  2. Probe Arrhenius (data_source: probe_tc)
    Aggregates probe_tc fits on the fly. You can filter by constructs, buffers, bases, reaction-group IDs, etc. Weighted regression uses log_kobs_err when available.

  3. Two-state melt (mode: two_state_melt)
    A placeholder stub exists; hook in your custom engine or supply series overrides until implemented.

Outputs

  • A run record per series in tempgrad_fit_runs (fit kind, scope, data source, metadata).
  • Tall-format parameters and diagnostics in tempgrad_fit_params.
  • JSON artifacts under <output_dir>/<label>/tempgrad_fit_latest/results/tempgrad_result.json.

Examples

# NMR degradation Arrhenius (species: dms)
nerd tempgrad_fit --config manuscript_pipeline/03_nmr_arrhenius/tempgrad_deg.yaml \
                  --db manuscript_pipeline/nerd.sqlite

# Probe Arrhenius, melted-only ATP nucleotides
nerd tempgrad_fit --config examples/probe_tempgrad_arrhenius/config.yaml \
                  --db examples/nerd.sqlite

# Manual series override
nerd tempgrad_fit --config configs/tempgrad_manual.yaml --db nerd.sqlite

After running tempgrad_fit, downstream notebooks or reports can summarize activation energies, global slope/intercept, and melt diagnostics directly from the tall-format tables.