Skip to main content
skill CX Codex

tooluniverse-dose-response — Claude Code skill for curve fitting

tooluniverse-dose-response fits a 4PL curve to concentration-response measurements and estimates IC50 or EC50. Use it to compare compound potency or assess how well a curve fits.

Apache-2.0 128 installs ★ 1,798 updated 2026-10-24

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add mims-harvard/ToolUniverse --skill tooluniverse-dose-response -a claude-code
Open on GitHub

What tooluniverse-dose-response does

The skill analyzes paired concentration and response measurements for inhibition or activation assays. Its four-parameter logistic model returns IC50 or EC50, Hill slope, Emax, Emin, and R². For two compounds, it calculates the IC50 shift and identifies which one is more potent.

Provide at least 4 data points, with concentrations on a linear scale and in consistent units. If the curve does not reach plateaus at both ends, the potency estimate is extrapolated and unstable. The 4PL model does not describe biphasic or bell-shaped responses.

For a single dataset, call tu run DoseResponse_calculate_ic50; DoseResponse_compare_potency compares two compounds. The scripts/fit_dose_response.py script fits a CSV using scipy. IC50 depends on assay conditions, so it is not an affinity estimate such as Ki without a Cheng-Prusoff correction.

What it helps with

  • estimate IC50 or EC50 from concentration-response data
  • compare potency between two compounds
  • check the quality of a fitted curve
  • retrieve Hill slope and Emax/Emin plateaus

Example prompts for the agent

  • Fit an IC50 curve to these concentrations and responses, then assess the fit quality.
  • Compare the potency of these two compounds from their concentration-response measurements and report the IC50 shift.

How to install tooluniverse-dose-response

1Skills CLI (Vercel)

npx skills add mims-harvard/ToolUniverse --skill tooluniverse-dose-response -a claude-code

2Manually: put the files in place

~/.claude/skills/tooluniverse-dose-response/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/tooluniverse-dose-response/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about tooluniverse-dose-response

How do I install tooluniverse-dose-response in Claude Code?

Run npx skills add mims-harvard/ToolUniverse --skill tooluniverse-dose-response -a claude-code. The skill file is at ~/.claude/skills/tooluniverse-dose-response/SKILL.md.

What does tooluniverse-dose-response do?

It fits a four-parameter logistic curve to concentration-response data and returns potency estimates and curve parameters, along with fit quality.

What data does the IC50 calculation require?

Provide at least 4 concentration-response pairs. Use linear concentrations and keep their units consistent across all points.

Source text

SKILL.md · Apache-2.0
---
name: tooluniverse-dose-response
description: Dose-response / concentration-response curve fitting — IC50, EC50, Hill slope, Emax/Emin efficacy, and relative potency from paired concentration vs response data (enzyme/cell assays, drug screening, agonist/antagonist pharmacology). Fits the 4-parameter logistic (Hill sigmoidal) model. Use when you have concentrations + responses and need a potency value, to compare two compounds' potency, or to judge curve quality. NOT for image-derived dose-response (use tooluniverse-image-analysis) and NOT for survival/regression (use tooluniverse-statistical-modeling).
disable-model-invocation: true
---

# Dose-Response / Concentration-Response Analysis

Turn paired **concentration vs response** measurements into a potency (IC50/EC50), a Hill slope, an efficacy (Emax), and a quality judgment — and compare potency between compounds.

## When to use this

- Enzyme inhibition / activation assays, cell viability, reporter assays, radioligand binding, agonist/antagonist pharmacology.
- You have a list of concentrations and the response at each, and want IC50/EC50 + Hill slope.
- You want to say "compound A is N-fold more potent than B."

The model is the 4-parameter logistic (4PL) / Hill sigmoidal:
`f(x) = Emin + (Emax − Emin) / (1 + (EC50/x)^n)` — where `n` is the Hill slope.

## Step 1 — Prepare the data (where results go wrong)

| Issue | What to do |
|---|---|
| **Concentration units** | Pick ONE unit (µM, nM, M) and use it for every point. The IC50 comes back in that unit. Don't mix. |
| **Log vs linear concentrations** | Pass concentrations on the **linear** scale (e.g. `0.01, 0.1, 1, 10`), not log10. The fitter logs internally. |
| **Zero/control concentration** | Drop a literal `0` concentration (log(0) is undefined). Keep it only as the Emin/Emax reference if normalizing. |
| **Direction** | Inhibition curves go high→low (IC50); activation curves go low→high (EC50). The tools handle both; just be consistent. |
| **Normalization** | Convert raw signal to % of control if you want Emax/Emin near 100/0: `% = 100 × (raw − blank)/(control − blank)`. Raw values also fit, but plateaus are then in raw units. |
| **Replicates** | Average technical replicates per concentration before fitting, or pass all points (the fit weights them equally). |

**Coverage requirement:** you need **≥4 points** (the tools require it) and ideally **6–8 spanning both plateaus** — points clearly above and clearly below the inflection. A curve that never plateaus gives an unreliable, extrapolated IC50 (see Step 4).

## Step 2 — Fit / get the potency

Single curve → IC50 or EC50 (same math; "IC50" for inhibition, "EC50" for activation):

```bash
tu run DoseResponse_calculate_ic50 '{"operation":"calculate_ic50",
  "concentrations":[0.001,0.01,0.1,1,10,100],
  "responses":[98,95,80,45,12,3]}'
```

Returns `ic50`, `ic50_95_confidence_interval`, `hill_slope`, `emax`, `emin`, `r_squared`, `log_ic50`.

Full 4PL parameters only → `DoseResponse_fit_curve` (same inputs). Two compounds → `DoseResponse_compare_potency` with `conc_a/resp_a/conc_b/resp_b` (returns each IC50 + `ic50_fold_shift_b_over_a` + `more_potent`).

For non-standard needs (constrained plateaus, weighting, plotting), `scripts/fit_dose_response.py` runs a scipy 4PL fit from a CSV and matches the tool.

## Step 3 — Interpret the four parameters

| Parameter | Meaning | Sanity check |
|---|---|---|
| **IC50 / EC50** | Concentration giving half-maximal effect — the potency. Lower = more potent. | Should fall *within* your tested range; if it's at/beyond an endpoint, the curve is incomplete (Step 4). |
| **Hill slope `n`** | Steepness / apparent cooperativity. ~1 = simple one-site. >1.5 = steep/positive cooperativity (or non-specific). <0.5 = shallow/multiple sites or heterogeneity. | A wildly large `|n|` (>4) usually means a bad fit or too few transition points, not real cooperativity. |
| **Emax** | Maximal response (top plateau) = efficacy. | For % data, full agonist ≈100; a **partial agonist** plateaus well below 100 even at saturating dose. |
| **Emin** | Bottom plateau (baseline/floor). | For % inhibition data, ≈0 for a complete inhibitor. |
| **r²** | Fit quality. | ≥0.95 good; <0.90 → inspect for outliers, wrong model, or incomplete curve before trusting the IC50. |

**Potency comparison:** report the **fold-shift** in IC50/EC50 (e.g. "A is 6.2× more potent than B"), and only call it meaningful if both fits are good (r²≥0.95) and the Hill slopes are comparable — a potency ratio between curves of very different slope is not a clean comparison.

## Step 4 — Quality gotchas (state these, don't hide them)

- **Incomplete curve / no plateau.** If responses don't flatten at both ends, Emax/Emin (and thus IC50) are extrapolated and unstable. Report the IC50 as "approximate / right-shifted of the tested range" and recommend wider concentrations.
- **<4–5 points or none near the inflection.** The fit can converge to a nonsense IC50 with a high r². Check that points actually bracket the IC50.
- **Biphasic / U-shaped data.** A single 4PL is wrong for hormesis or two-site behavior — the fit will look poor (low r²); flag it rather than forcing one IC50.
- **IC50 vs Ki.** IC50 depends on assay conditions (substrate/ligand concentration). Don't report IC50 as an affinity (Ki) without a Cheng-Prusoff correction.
- **Units.** The IC50 is only as correct as the concentration unit you fed in — always state the unit.

## Honest limitations

- 4PL assumes a monotonic sigmoid; it cannot describe biphasic, bell-shaped, or steep all-or-none responses.
- A confident IC50 from a poor or incomplete curve is the most common error — let r² and curve coverage gate how you report it.
- Potency (IC50/EC50) is not efficacy (Emax) — a more potent compound can be a weaker (partial) agonist; report both.

## Related skills
- `tooluniverse-image-analysis` — dose-response on image-derived measurements (.tif, colony, fluorescence).
- `tooluniverse-gpcr-structural-pharmacology` / `tooluniverse-network-pharmacology` — receptor pharmacology context.
- `tooluniverse-statistical-modeling` — general regression, EC50 via spline, power analysis.

© mims-harvard · Apache-2.0 · mims-harvard/ToolUniverse

Similar skills

SKILL.md skills catalog →

From the same set136

The whole set →
tooluniverse-residue-functional-mechanism-interpretation The tooluniverse-residue-functional-mechanism-interpretation skill for Claude Code explains the significance of protein residues using structural context, UniProt annotations, and optional DMS or SAE evidence. CX Codex +1 140installs tooluniverse-mendelian-randomization The tooluniverse-mendelian-randomization skill tests whether one factor causally drives another (e.g., cholesterol → heart disease) using genetic variants as instrumental variables. CX Codex +1 137installs tooluniverse-primer-design tooluniverse-primer-design designs PCR and qPCR primers, computes Tm and Ta, and screens oligonucleotides for hairpins and dimers. CX Codex +1 131installs tooluniverse-diagnostic-test-evaluation tooluniverse-diagnostic-test-evaluation evaluates diagnostic test and biomarker accuracy from 2×2 tables or continuous scores. Computes sensitivity, specificity, ROC AUC, and Bayesian post-test probability. CX Codex +1 129installs tooluniverse-enzyme-kinetics tooluniverse-enzyme-kinetics fits Michaelis-Menten enzyme kinetics from substrate-velocity experimental points and identifies inhibition modes. CX Codex +1 128installs tooluniverse-drug-synergy tooluniverse-drug-synergy quantifies whether two drugs act synergistically, additively, or antagonistically based on measured single-agent and combination effects. CX Codex +1 127installs tooluniverse-clinical-risk-scoring tooluniverse-clinical-risk-scoring — a skill for Claude Code, Codex, Cursor and other agents that computes validated clinical risk scores from individual patient data. CX Codex +1 125installs tooluniverse-product-safety-surveillance tooluniverse-product-safety-surveillance is a skill that searches openFDA for adverse events, recalls, and safety reports on medical devices, supplements, cosmetics, veterinary drugs, and medication shortages. CX Codex +1 123installs

Project sponsors

Companies that keep this analytics open