Parameter inference of a basic p53 model using ABC
Eszter Lakatos and Michael Barclay
Group meeting 29th October 2014
Background
Study p53 reaction to cellular stress on single cell level
Compare and model (stochastically) normal and drug-treated cells
MCF-7 cells
human breast cancer cell line
BE cells
bigger cells, faster growth
Actinomycin D
polypeptide antibiotic with anti-cancer activity transcriptional inhibitor
Experimental results
0 1.5 3 4.5 6 x 104 5 10 15 20 80nM ActinomycinD (n = 65) 0 1 2 3 4 5 x 104 5 10 15 20 25 40nM ActinomycinD (n = 58) 0 4 8 12 x 105 4 8 12 16 BE cells (n = 51) 0 0.4 0.8 1.2 1.6 2 x 104 5 10 15 20 25 MCF7 cells (n = 45) 0 0.75 1.5 2.25 3 x 104 10 20 30 40 8nM ActinomycinD (n = 69)Model
˙ m=k1−k2m ˙ p =k3m−k4p k1 - transcription k2 - mRNA degr. k3 - translation k4 - protein degr.Model
˙ m=k1−k2m ˙ p =k3m−k4p k1 - transcription k2 - mRNA degr. k3 - translation k4 - protein degr.Combined model of different cell populations
MCF7: ˙ m1=k1−k2m1 ˙ p1=k3m1−k4p1 k1 - transcription k2 - mRNA degr. k3 - translation k4 - protein degr. BE: ˙ m2 =k1s−k2sm2 ˙ p2=k3sm2−k4sk5p2 s - scaling factor k5 - modifier Hypothesis: k5<<1Combined model of different cell populations
8nM ActD: ˙ m1 =k1−k2m1 ˙ p1 =k3m1−k4p1 k1 - transcription k2 - mRNA degr. k3 - translation k4 - protein degr. 40nM ActD: ˙ m2 =k1−k2m2 ˙ p2 =k3m2−k5p2 k5 - protein degr. 80nM ActD: ˙ m3 =k1−k2m3 ˙ p3 =k3m3−k6p3 k6 - protein degr. Hypothesis: k6 <k5<k4Inference using ABC
Algorithm
Sample parameters from (prior) distribution
Simulate each particle n times: SDE
Compare experimental and simulated populations
Accept particle if distance<
Inference using ABC
Algorithm
Sample parameters from (prior) distribution
Simulate each particle n times: SDE
Compare experimental and simulated populations
Accept particle if distance<
Iterate
Conclusions from the first months:
Kolmogorov-Smirnov distance
Non-identifiability
Rescale parameters by settingk2=1
Check the limitations with simulated experimental data
Results I. - Identifiability
Populations (n=60) simulated with the same parameters
0 5000 10000 15000 20000 25000 0.0 0.2 0.4 0.6 0.8 1.0 sim. experimental simulation0 simulation1 0 5000 10000 15000 20000 0 5 10 15 2025 3035
Results I. - Identifiability
Results I. - Identifiability
Results I. - Identifiability
Results I. - Identifiability
Combined model of different cell populations
MCF7: ˙ m1=k1−k2m1 ˙ p1=k3m1−k4p1 k1 - transcription k2 - mRNA degr. k3 - translation k4 - protein degr. BE: ˙ m2 =k1s−k2sm2 ˙ p2=k3sm2−k4sk5p2 s - scaling factor k5 - modifier Hypothesis: k5<<1Combined model of different cell populations
8nM ActD: ˙ m1 =k1−k2m1 ˙ p1 =k3m1−k4p1 k1 - transcription k2 - mRNA degr. k3 - translation k4 - protein degr. 40nM ActD: ˙ m2 =k1−k2m2 ˙ p2 =k3m2−k5p2 k5 - protein degr. 80nM ActD: ˙ m3 =k1−k2m3 ˙ p3 =k3m3−k6p3 k6 - protein degr. Hypothesis: k6 <k5<k4Results II. - Real data
Results II. - Real data
Combined model of MCF7 and BE cells
0 5000 10000 15000 20000 0.0 0.2 0.4 0.6 0.8 1.0 0 200000 400000 600000 800000 1000000 1200000 0.0 0.2 0.4 0.6 0.8 1.0 simulation1 simulation2 experimental
Results II. - Real data
Results II. - Real data
Combined model of 3 doses of ActinomycinD treatment
0 5000 10000 15000 20000 25000 30000 0.0 0.2 0.4 0.6 0.8 1.0 0 10000 20000 30000 40000 50000 0.0 0.2 0.4 0.6 0.8 1.0 0 10000 20000 30000 40000 50000 60000 70000 0.0 0.2 0.4 0.6 0.8 1.0 simulation1 simulation2 experimental
Results II. - Real data
Combined models of 3 doses of ActinomycinD treatment Model 1: ActD affects protein degradation
Model 2: ActD affects transcription Model 3: ActD affects translation
Results II. - Real data
Combined models of 3 doses of ActinomycinD treatment
0 5000 10000 15000 20000 25000 30000 0.0 0.2 0.4 0.6 0.8 1.0 0 10000 20000 30000 40000 50000 60000 0.0 0.2 0.4 0.6 0.8 1.0 0 20000 40000 60000 80000 100000 120000 0.0 0.2 0.4 0.6 0.8 1.0 simulation13 simulation83 simulation98 experimental
Plans, problems, questions
More data (faster set-up)
Fluorescent data from MCF7 and BE cells
Both p53 and Mdm2 labelling
Low molecular numbers of protein are not measured reliably
E.g. in the MCF7 cell line, about 10 cells were thrown out
Can be considered in distance
Two populationsof cells