Table 1 ∣.

Summary of modeling objectives and candidate models for each data set¹.

	Model A			Model B			Model C			Model D
Additional mechanism→ Modeling objective ↓	Base case			Cas13a-gRNA deactivates over time			Negative relationship between transcription rate and T7 RNAP			Non-monotonic relationship between RNase H cleavage rate and RNase H
	Data set 1	Data set 2	Data set 3	Data set 1	Data set 2	Data set 3	Data set 1	Data set 2	Data set 3	Data set 1	Data set 2	Data set 3
Objective 1: Time course trajectories have a sigmoidal shape	Yes	Yes2	Yes2	Yes	Yes2	Yes2	Yes	Yes	Yes	Yes	Yes	Yes
Objective 2: Plateaus in readout can occur at various times depending on the condition	Yes	Yes2	Yes2	Yes	Yes2	Yes2	Yes	Yes	Yes	Yes	Yes	Yes
Objective 3: The magnitude of the plateau can vary depending on the condition	No	No	No	Yes	No	No	Yes	Yes	Yes	Yes	Yes	Yes
Objective 4: Increasing RT to a relatively high concentration increases the readout	No	No	Yes	Yes	Yes	Yes	Yes	No	Yes	Yes	Yes	Yes
Objective 5: Increasing T7 RNAP to a relatively high concentration decreases the readout	No	No	No	No	No	No	Yes	Yes	Yes	Yes	Yes	Yes
Objective 6: Increasing RNase H to a relatively high concentration increases, then decreases the readout	No	No	No	No	No	No	No	No	No	No3	Yes	Yes
Quantitative agreement (R2)	0.43	0.26	0.31	0.53	0.27	0.33	0.99	0.87	0.78	0.99	0.96	0.85
Quantitative agreement (MSE)	0.062	0.040	0.051	0.045	0.037	0.048	0.0021	0.0066	0.018	0.0019	0.0026	0.016

¹Each column is an additional mechanism added to the model. For example, Model C includes mechanisms from Models A and B. ‘Yes’ indicates that a version met an objective, and ‘No’ indicates that it did not. Calibration and analysis of suboptimal candidate models are described in Supplementary Fig. 11. Quantitative agreement is reported as the MSE or R² between experimental data and simulated data for the subset that was used for training in Fig. 4 (Supplementary Fig. 6b).

²The determination of whether these objectives were met for Data Sets 2 and 3 is based only on the quantitative metrics from the Hill fit (i.e., a high R² value between the trajectories and Hill fits and a range of t_1/2 values) (Supplementary Note 3).

³Although this modeling objective was not met for Data Set 1, given the experimental error in the RNase H sweeps it is unclear whether this data set has a non-monotonic relationship between RNase concentration and readout, which is indicative of potential error in the amount of each reagent dispensed from the liquid handling robot for each experiment.