Mory ReactionsFigure 5. The p53-MDM2 core module. P53 protein is activated by the upstream signal (represented by ATM kinase) and form tetramers as the TFs. p53 positively regulates gene MDM2 by activating its transcription, whereas MDM2 negatively regulates p53 by promoting its ubiquitination and degradation. Regulatory mechanisms for the expression of gene MDM2 follow the same assumptions in Figure 1, which are characterized by the two memory windows for the continuous transcription and inactivity time periods of gene MDM2. doi:10.1371/journal.pone.0052029.gFigure 6. Stochastic Potassium clavulanate simulations of the p53-MDM2 core module. The upstream signal represented by the ATM kinase activities (measured from Fig. 1 in [50]) has two pulses in (A) or four pulses in (D). Five simulations of the p53 copy numbers based on two pulses (B) and four pulses (E) of the upstream signal; and the corresponding MDM2 copy numbers in five simulations induced by two pulses (C) and four pulses (F) of p53 activities. doi:10.1371/journal.pone.0052029.gModeling of Memory ReactionsFigure 7. Damped oscillation of the p53 module in a population of cells. (A) Fractions of cells showing different pulse numbers of ATM activity when cells were irradiated by different gamma doses. The averaged copy numbers of p53 (B) and MDM2 (C) based on 1000 simulations. (Solid-line: gamma dose 10 Gy, dash-dot-line: 2.5 Gy, and dash-line: 0.3 Gy). doi:10.1371/journal.pone.0052029.gthe firing of memory reactions also depends on the competition with other reactions if it is within the memory time period. In addition, the key feature of delayed reaction is the time difference between the firing of a chemical reaction and manifest of its products. However, the products of a memory reaction are generated immediately after its firing. In this work we also proposed the delayed memory reaction if the reaction is conditional to the path of memory events as well as there is delay between the firing of the chemical reaction and manifest of its products. Furthermore, molecules involving in delayed reactions are static during the delayed time period because they are reserved for the product manifest in a future time point; however, molecules involving in memory 
reactions are dynamic since they involve in other reactions in the memory window. Thus the memory and time delay are two distinct features of chemical reactions, though these two types of reactions are connected to a fixed length of time period. Regarding the necessary of memory reactions, one may argue that the memory phenomena may be simply SMER28 realized by using additional species and additional chemical reactions within the classic SSA framework. If this modeling scheme were implemented without using memory reactions, the competitive nature of the elementary stochastic chemical reactions would cause that the time period of a particular biological/cellular event does not follow the distribution observed in experiments. For example, the rapid reinitiation rate of transcription should be matched by a large termination rate of gene expression, namely the rate of TFdisassociating from the DNA promoter site. In this case the exit strategy of gene expression is realized by the competitive reaction of TF disassociation. However, our simulation results suggested that it is difficult to use this strategy to realize the relatively constant time periods of gene expression that were observed in experiments. In this work we proposed the memory reaction to realize s.Mory ReactionsFigure 5. The p53-MDM2 core module. P53 protein is activated by the upstream signal (represented by ATM kinase) and form tetramers as the TFs. p53 positively regulates gene MDM2 by activating its transcription, whereas MDM2 negatively regulates p53 by promoting its ubiquitination and degradation. Regulatory mechanisms for the expression of gene MDM2 follow the same assumptions in Figure 1, which are characterized by the two memory windows for the continuous transcription and inactivity time periods of gene MDM2. doi:10.1371/journal.pone.0052029.gFigure 6. Stochastic simulations of the p53-MDM2 core module. The upstream signal represented by the ATM kinase activities (measured from Fig. 1 in [50]) has two pulses in (A) or four pulses in (D). Five simulations of the p53 copy numbers based on two pulses (B) and four pulses (E) of the upstream signal; and the corresponding MDM2 copy numbers in five simulations induced by two pulses (C) and four pulses (F) of p53 activities. doi:10.1371/journal.pone.0052029.gModeling of Memory ReactionsFigure 7. Damped oscillation of the p53 module in a population of cells. (A) Fractions of cells showing different pulse numbers of ATM activity when cells were irradiated by different gamma doses. The averaged copy numbers of p53 (B) and MDM2 (C) based on 1000 simulations. (Solid-line: gamma dose 10 Gy, dash-dot-line: 2.5 Gy, and dash-line: 0.3 Gy). doi:10.1371/journal.pone.0052029.gthe firing of memory reactions also depends on the competition with other reactions if it is within the memory time period. In addition, the key feature of delayed reaction is the time difference between the firing of a chemical reaction and manifest of its products. However, the products of a memory reaction are generated immediately after its firing. In this work we also proposed the delayed memory reaction if the reaction is conditional to the path of memory events as well as there is delay between the firing of the chemical reaction and manifest of its products. Furthermore, molecules involving in delayed reactions are static during the delayed time period because they are reserved for the product manifest in a future time point; however, molecules involving in memory reactions are dynamic since they involve in other reactions in the memory window. Thus the memory and time delay are two distinct features of chemical reactions, though these two types of reactions are connected to a fixed length of time period. Regarding the necessary of memory 
reactions, one may argue that the memory phenomena may be simply realized by using additional species and additional chemical reactions within the classic SSA framework. If this modeling scheme were implemented without using memory reactions, the competitive nature of the elementary stochastic chemical reactions would cause that the time period of a particular biological/cellular event does not follow the distribution observed in experiments. For example, the rapid reinitiation rate of transcription should be matched by a large termination rate of gene expression, namely the rate of TFdisassociating from the DNA promoter site. In this case the exit strategy of gene expression is realized by the competitive reaction of TF disassociation. However, our simulation results suggested that it is difficult to use this strategy to realize the relatively constant time periods of gene expression that were observed in experiments. In this work we proposed the memory reaction to realize s.
