When you run nonlinear regression from a Multiple Variables data table, the first tab of the analysis parameters dialog is the Data tab. This is where you tell Prism which columns contain your X values (predictor), which contain your Y values (response), and which define the separate datasets you want to fit with individual curves.

Overview of Variable Assignment

The Data tab has two main sections. On the left, you'll see all the variables available for the analysis from the source data table. On the right are three boxes where you assign variables to their respective roles in the analysis.

To assign a variable, you can either drag it from the left side into the appropriate box on the right, or click the "+ Add variable..." button in each assignment box and select from the list. Once assigned, you can remove a variable by hovering over it and clicking the X button that appears.

Response (Y) Variable(s)

The response variable is your measured outcome. This is the dependent variable that you expect to change as a function of X. In dose-response experiments, this might be percent inhibition, cell viability, or receptor binding. In enzyme kinetics, it might be reaction velocity. In pharmacokinetics, it might be drug concentration in plasma.

The response variable must be continuous numeric data. Prism will fit curves to predict Y from X, so your Y values need to be quantitative measurements.

Assigning multiple response variables

One powerful feature of Multiple Variables tables is that you can assign multiple response variables measured at the same X values. This is common when you measure multiple readouts in the same experiment. For example, viability and apoptosis at each drug concentration, or multiple fluorescence channels in a binding assay.

When you assign multiple response variables, Prism fits separate curves for each one. If you also assign grouping variables, you'll get curves for every combination of response variable and group. So with 2 response variables and 3 treatment groups, you'd get 6 fitted curves (2 × 3 = 6).

Requirements for response variables:

•Must be continuous numeric data

•All values must use consistent units

•Can include replicates (multiple Y values at the same X value)

Good examples of response variables:

•Percent_Inhibition

•Cell_Viability

•Fluorescence_Intensity

•Binding_Signal

•Enzyme_Velocity

•Absorbance

•Luminescence

•Response_Percent

•Plasma_Concentration

Predictor (X) Variable

The predictor variable is your independent variable. This is the factor you're systematically varying to see its effect on the response. In dose-response experiments, this is drug concentration or dose. In enzyme kinetics, it's substrate concentration. In time course experiments, it's time. In binding studies, it's ligand concentration.

The X variable must be continuous numeric data. Prism uses these X values to fit your chosen equation and predict the relationship between X and Y.

Requirements for the predictor variable:

•Must be continuous numeric data

•All values must use consistent units

•Typically need at least 5-6 distinct X values that span the range of the curve for reliable curve fitting

•Can only assign one predictor variable per analysis

Good examples of predictor variables:

•Concentration_uM

•Dose_mg_per_kg

•Time_hours

•Substrate_Concentration_mM

•Ligand_nM

•Log_Concentration

•Temperature_Celsius

The column name you use will automatically appear as the X-axis label on your fitted curve plots, so use descriptive names that include units.

A note about log-transformed X values

For dose-response curves spanning multiple orders of magnitude (like 0.001 to 100 µM), you may have your X values already log-transformed in your data table. That's fine - just make sure you're using an appropriate equation for log(X) rather than X. Alternatively, you can store the actual concentrations and let Prism's dose-response equations handle the log transformation internally, which is often more intuitive.

Grouping Variable(s)

Grouping variables are optional, but they're essential when you want to fit separate curves for different experimental conditions. Common examples include different compounds in a screening experiment, different cell lines, different treatment conditions, different time points, or different batches.

When you don't assign any grouping variables, Prism fits a single curve to all your data. When you assign one or more grouping variables, Prism fits separate curves for each unique combination of group levels, giving you individual parameter estimates (like EC50, Vmax, Kd, etc.) for each curve.

Assigning multiple grouping variables

You can assign multiple grouping variables to fit curves for combinations of factors. For example, you might group by both Compound (5 test compounds) and Cell_Line (3 different cell lines). Prism will fit 15 separate curves (5 × 3 = 15), one for each compound-cell line combination.

This is powerful for large-scale experiments, but keep these considerations in mind:

•More grouping variables means more curves to fit

•Each curve needs adequate data (typically at least 5-6 X values with at least one Y value per X)

•Results tables and graphs can become large with many curves

•More complex models with more parameters may require more data

Requirements for grouping variables:

•Must be categorical variables

•Can have two or more levels

•Use consistent spelling and capitalization for group labels

Good examples of grouping variables:

•Compound (Drug_A, Drug_B, Drug_C, Control)

•Cell_Line (HeLa, A549, HEK293, MCF7)

•Treatment (Vehicle, Pretreated, Co-treated)

•Timepoint (4hr, 24hr, 48hr, 72hr)

•Batch (Batch_1, Batch_2, Batch_3)

•Genotype (WT, Heterozygous, Knockout)

•Temperature (25C, 37C, 42C)

Example Variable Assignments

Let's look at a few common scenarios and how you'd assign variables for each.

Scenario 1: Simple dose-response curve (single compound)

You're fitting a dose-response curve to a single compound with no groups to compare.

Variable assignment:

•Response (Y): Percent_Inhibition

•Predictor (X): Concentration_uM

•Grouping: None

Result: One fitted curve with estimated EC50, Hill slope, and other parameters.

Scenario 2: Comparing multiple compounds

You're comparing dose-response curves for several test compounds.

Variable assignment:

•Response (Y): Response

•Predictor (X): Dose_uM

•Grouping: Compound with values Drug_A, Drug_B, Drug_C, Control

Result: Four fitted curves (one per compound) with individual parameter estimates for each, allowing you to compare EC50 values across compounds.

Scenario 3: Multiple readouts from the same experiment

You measured both viability and apoptosis at each drug concentration.

Variable assignment:

•Response (Y): Viability AND Apoptosis (both assigned)

•Predictor (X): Concentration_uM

•Grouping: Compound with values Drug_A, Drug_B

Result: Four fitted curves:

•Viability-Drug_A

•Viability-Drug_B

•Apoptosis-Drug_A

•Apoptosis-Drug_B

This lets you compare how the same compounds affect different cellular endpoints.

Scenario 4: Multi-factorial experiment

You tested multiple compounds across different cell lines to see if potency varies by cellular context.

Variable assignment:

•Response (Y): Viability_Percent

•Predictor (X): Concentration_uM

•Grouping: Compound AND Cell_Line (both assigned)

Result: If you have 3 compounds and 2 cell lines, you get 6 fitted curves (3 × 2 = 6) representing all combinations, allowing you to compare EC50 values across both compounds and cell types.

Scenario 5: Enzyme kinetics

You're measuring enzyme velocity at different substrate concentrations to estimate Km and Vmax.

Variable assignment:

•Response (Y): Velocity_uM_per_min

•Predictor (X): Substrate_mM

•Grouping: Enzyme_Prep (to compare different preparations or mutants)

Result: Separate Michaelis-Menten curves for each enzyme preparation, with Km and Vmax estimates for each.

How to Assign Variables

There are two ways to assign variables to their roles:

Method 1: Drag and drop

1.Click and hold on a variable name in the Available variables list

2.Drag it to the appropriate assignment box on the right

3.Release to drop it into the box

Method 2: Add button

1.Click the "+ Add variable..." button in any assignment box

2.Select or search for the variable you want to assign from the list that appears

To remove a variable:

•Hover over the variable, then click the X button next to the variable name in the assignment box

•The variable will return to the Available variables list

Common Questions

Q: How many data points do I need per curve?

A: This depends on the complexity of your equation. Simple two-parameter equations might work with 4-5 points, but most dose-response curves (4-parameter logistic) need at least 5-6 distinct X values with good coverage of the curve's range. More complex equations need more data points. Replicates at each X value improve precision but don't substitute for having adequate range of X values.

Q: Can I assign multiple predictor (X) variables?

A: No, nonlinear regression fits Y as a function of a single X variable. If you have multiple predictors, you'd need multiple regression or other multivariate analysis methods.

Q: What happens if I assign both multiple response variables AND multiple grouping variables?

A: Prism will fit curves for every combination. With 3 response variables and 4 groups, you'll get 12 curves (3 × 4 = 12). This is useful for comprehensive analysis but make sure each curve has adequate data.

Q: How does Prism handle rows with missing values?

A: Prism automatically excludes any row that has a missing value in the response variable, predictor variable, or any assigned grouping variable. Only complete cases are used for curve fitting.

Q: My data has replicates. Do I need to average them first?

A: No, leave replicates as separate rows. Prism will use all data points when fitting each curve, which is statistically more appropriate than pre-averaging. Replicates at the same X value help define the variability in your data.

Q: Can I fit different equations to different curves?

A: No, in a single analysis, Prism fits the same equation to all curves defined by your grouping variables. If you need to fit different equations, run separate analyses for each subset of data.

Q: Do I need to sort my data in any particular order?

A: No, row order doesn't matter. Prism will organize the data as needed for curve fitting and plotting.