Code for "Estimating mutation rates in a Markov branching process using approximate Bayesian computation"

This software package includes the source code (mostly in MATLAB) for our manuscript "Estimating mutation rates in a Markov branching process using approximate Bayesian computation". There are three parts in the package: I. ABC-based estimator and MOM/MLE in this main folder, II. Simulation study 1 (for the constant mutation scenario) and Simulation study 2 (for the piece-wise constant mutation scenario) under subfolder simulation/, and III. Real data analysis under subfolder real/.

Estimating a constant mutation rate

• ABC-based estimator with "exact" simulator:
  • fluc_exp1.m: The "exact" simulation. Generate fluctuation data in parallel cultures from bMBP model with constant mutation rate. Used everywhere.
  • ABC_fluc_exp1.m: Estimate mutation rate by ABC for fluctuation data with constant mutation rate. Used in simulation/simu1gendata_orisim.m and simulation/simu1_orisim_statlnx9.m.
  • trainGPS.m: Train GPS model for fluctuation data with constant mutation rate. Used in ABC_fluc_exp1.m, also in demoGPS_fluc_exp1_rev.m.
  • tnrnd.m: Generate random sample from truncated normal. Used in ABC_fluc_exp1.m.
  • mut_bMBP.m: Generate (z, x) data in a single culture from bMBP model with constant mutation rate. Used in fluc_exp1.m.
  • selsummary_fluc_exp1.m: Compare the response curves of three different summary statistics (Fig. 1).
  • demoGPS_fluc_exp1_rev.m: Demonstrate the performance of GP regression (Fig. 2).
• ABC-based estimator with "fast" simulator
  • fluc_exp1_rev.m: The "fast" simulator. Generate fluctuation data in parallel cultures from approximated bMBP model with constant mutation rate. Differential growth between mutants and normal cells is allowed. Used everywhere.
  • ABC_mu1.m: GPS-ABC estimator based on constant mutation rate assumption. Used in simulation/simulation1_server.m and real/werngren_server1.m.
  • ABC_MCMC.m: ABC-MCMC estimator based on constant mutation rate assumption. Used in simulation/simulation1_MCMC_server.m.
  • ABC_mu1a.m: GPS-ABC estimator to estimate constant mutation rate and relative growth rate of mutants vs. normal cells. Used in real/werngren_server1a.m.
  • get_IniX1*.m: Generate initial training samples for fitting GP surrogate model. Used in ABC_mu1*.m.
  • get_theta1.m and get_theta1a.m: Propose from the proposal distribution. Used in ABC_mu1.m and ABC_mu1a.m, respectively.
  • tnorm.m: Generate random sample from truncated normal. Used in ABC_mu1.m, ABC_mu1a.m, ABC_mu2.m and ABC_mu2a.m.
  • unconditionError: Calculate unconditional error of making an accept/reject decision. Used in ABC_mu1.m, ABC_mu1a.m, ABC_mu2.m and ABC_mu2a.m.
  • mut_bMBP_rev.m: Generate approximated (z, x) data in a single culture based on the conditional distribution of the occurring time of mutations.
• MOM/MLE estimator:
  • MOMMLE_fluc_exp1.m: Estimate mutation rate by MOM/MLE for fluctuation data with constant mutation rate. Used in simulation/simulation1andsimulation/simu1_orisim_statlnx9.m`.

Estimating parameters in a piece-wise constant mutation rate function

• ABC-based estimator with "exact" simulator:

  Functions and detailed instructions could be found in the subfolder exact simulator/.

• ABC-based estimator with "fast" simulator

  • fluc_exp2_rev.m: The "fast" simulator with a piece-wise constant mutation rate. Generate fluctuation data in parallel cultures . Differential growth between mutants and normal cells is allowed. Used everywhere.
  • ABC_mu2a.m: GPS-ABC estimator to esimate four parameters in a piece-wise constatnt mutation rate function, p1, p2, transition time τ and relative growth rate δ. Used in real/werngren_server2a.
  • ABC_mu2.m: GPS-ABC estimator to estimate three parameters in a piece-wise constant mutation rate function, p1, p2 and tarnsition time τ.
  • get_IniX2.m and get_Inix2a.m: Generate initial training samples for fitting GP surrogate model. Used in ABC_mu2.m and ABC_mu2a.m, respectively.
  • get_theta2.m and get_theta2a.m: Propose from the proposal distribution. Used in ABC_mu2.m and ABC_mu2a.m, respectively.
  • mut2stage_bMBP_rev.m: Generate approximated (z, x) data in a single culture based on the conditional distribution of the occurring time of mutations.

Note: The detailed model specifications are described as comments in the script.