 Good morning everyone, my name is Vivek Kumar Dhairi, I am a student of IEL Delhi. This is the Indian Institute of Technology Delhi. It was set up as part of the Indian Institute of Technology Act of 1971. It was in a new stream to create an army of scientists and engineers in India to lead the progress in their respective fields. I am a student of Department of Biomedical Engineering and Biotechnology at IEL Delhi. This summer I have been working in Dr. Reed's lab, analysing the effect of thermodynamic constraints on predictive capabilities of genome-sealed model of E.coli. In the particular model I have been working with is known as the IJR904 model. I will be telling you about that. So, modeling system biology is important because it can answer a lot of questions. We can try and interpret observed behaviours of system and answer questions such as why do cells make certain by-products. We can generate hypotheses about unexplained phenomena, how can cells utilize certain substrates and ask these questions. We can predict cellular behaviours of a developed system and predict how fast mutants can grow. And one of the most important uses is identifying metabolic engineering strategies answering questions such as which genes should be deleted to improve the production of a certain product, such as RFMR. So, modeling system biology is basically four steps. First we gather the information about the important components and the component interactions and biological networks from literature. This is then organised and assembled at the systems level using either textual, graphical or mathematical representation. And then we go on to the modelling. So, we convert the reconstruction into a model by reducing variables and equations. Based on the chemical and physical principles that we came across in the systems level. And once the model is complete, we then analyse results and we then compare it with existing experimental data. We use the results of the model to design experimental groups. And once the results are validated, then the whole cycle is complete. The results are again introduced. So, how do we go about network assembly and representation? So, let's say you have the glycolysis pathway and you have the reactant sex, echinacea, phosphorylisterin, sorry, glucose-exphosphorylisomerase, and all such reactions in the glycolysis pathway. This is arranged in the form of a matrix. And on the left, you can see all the metabolites. And at the bottom are the reactions. This matrix is known as the S matrix in our language. And you see the coefficients of the reactions. So, say if you have a minus one, that means one molecule of that particular metabolite is being consumed. And a plus one would mean a single molecule of that metabolite, echinacea. On the right here, you can see the genes associated with the enzymes and those associated with the reactions that they carry out. And further on the right is the glycolysis pathway in the form as it would appear in a network. So, suppose we have this flux through lacylendryhyde, free phosphate dehydration. And we want to switch that off. We can do that by switching between expression. So, how do we go about analyzing these metatonic networks? We do optimization through linear programming. And in optimization, we have an objective function, a function that is maximized or minimized by the optimal solutions. And in our case, we would have a function, we would either maximize the flux or a function of the flux. So, and we also apply constraints. And this is done to eliminate the solutions which would not make sense, which are infeasible, while leaving the selenium behaviors. So, we get a result like this, where the yellow space represents the feasible solutions. And the red dot is the optimal solution for the problem that would be specified. So, what are the constraints we place on metatonic networks? Well, the first constraint, that is a steady state mass balance constraint, is placed on all metatonic networks because we analyze these networks in the steady state. And suppose we have this metatonic light, which is generated through fluxes V1, V2 and V3. And we also have a flux that consumes this metatonic, the flux V4. Then the amount in which it is generated should be equal to the amount in which it is consumed. And that is basically our mass balance constraint. And this is applied to all the metatonic. So, that is the picture showing all the metatonic and how this particular constraint would be applied. Second type of constraint that is applied is an enzyme capacity constraint. So, we limit the flux within bounds, alpha and beta being the lower and upper bounds in this case. And third type of constraint that we apply is a thermodynamic constraint in which we basically fix the fluxes of all the irreversible reactions greater than zero. What are we missing out on? Well, we don't consider kinetics. And except for the thermodynamic constraints which I will be coming to, we don't consider concentrations. So, it's a steady state analysis and we do not look at these kind of two things. So, the thermodynamic constraint. Well, from the second law of thermodynamics, we know that delta G is equal to delta... Right, from second law of thermodynamics, we know that the flux to be positive for reaction, the delta G has to be negative. And we go back calculating the delta G using this formula. So, we have delta G is equal to delta G0, which is the delta GF reaction of stomach conditions. And we include the equilibrium constant term, which is... Say we have a reaction like this, and the equilibrium constant becomes this. And substituting that in the formula, we get delta G for the reaction. So, if the delta G is positive, then the flux is negative. And if the delta G is negative, then the flux has to be positive. So, this is the thermodynamic constraint that we apply. So, now we have two models that we use. The FBA model is the model that has been used up till now to look at the fluxes in the target pathways. And what I was trying to do was to apply thermodynamics instead of using the FBA model. And I will list out the differences in the two models. So, in case of FBA, the reversibility is decided on the basis of literature. So, we go into literature and find out which reactions are known to be reversible. Whereas, in the case of the TAP model, which is thermodynamic constraint model, the reversibility is decided on the basis of the delta G. So, as we decide that delta G is positive, then the reaction has to be in the reverse direction. And if delta G is negative, it has to be in the forward direction. In case of the only fluxes are important, our considerations have no role. In case of the thermodynamic model, considerations come into the picture because of the equilibrium constant. So, in that is considered. In case of FBA, the constraints include mass balance and reversibility because we are deciding prior to analyze the analysis that these particular reactions are reversible. In case of TFBA, mass balance is there. We place bounds on the higher and lower concentrations of thermodynamics. And we also include the thermodynamic constraint where delta G is positive, flux is negative, delta G is negative, flux is positive. So, I would like to present the two models in form of a superset subset situation. So, you have the TFBA which becomes a superset of the FBA because in case of the FBA, only certain reactions were known to be reversible and they were allowed to be reversible. Whereas, in case of TFBA, theoretically all reactions are reversible. It is the delta G which is then decided in the direction of the reaction. And part of that you have mass balance and thermodynamic constraint which I mentioned. In case of FBA, only mass balance is the constraint. Now, one of the published works that I was looking at during my stay here was actually focusing on this thermodynamic constraints on FBA. And their main focus was to remove cycles that might exist in this one, in the FBA model. So, but it would not give you an idea. It would still not allow some reactions which might be thermodynamically feasible to run in the reverse direction because the thermodynamic constraints are only placed on the reverse reactions. So, my work was to look at this one. So, the objectives of my project. Well, the first one was to make the thermodynamic model more accurate in the sense that it was giving a higher, it was predicting a higher growth rate than is that in the actual case. So, I was trying to place certain constraints which I'll be discussing to bring down the predicted growth rate. The second was to compare the prediction capabilities of FBA and TFBA for gene number three and five. The third part was to identify possible reversible reactions that are shown to be reversible in TFBA. But might not be known to be reversible in literature. So, we can then go on and find out if these are actually reversible. So, for the first part, well, I told you that for such a reaction, the equilibrium constant is this and the g is calculated using this formula. But this becomes, in fact, in some cases, where transport of metabolites is involved across the membrane. Because then the transport term also has to be included. And this transport term, it has two components. One is even the electrochemical gradient. And the second, you do the BHV. And so, which I calculated using this formula. So, now, when we do the, when we apply some random constraints, we get much better results. So, coming to the comparison part, the red bars are for FBA and the blue bars are for TFBA. And we see very similar results for the two, which means that we know most of the reactions are reversible to viso. There might be very few which might not be viso, which we have identified, which I'll be discussing in the next part. So, this was growth and blue cores. And we see that, so, g corresponds to growth and g corresponds to no growth. So, we compare model with the experiment. So, TFBA and FBA show very similar results. And the same is the case when we do the run on the glycerol. So, both models are pretty similar. So, these are the cases in which TFBA predicted much better. When we looked out these genes, we saw that TFBA gives growth growth phenotype, whereas your FBA model would show no growth when there's actually growth taking place in the experiment. So, let's analyze one case where this v0928 gene was knocked out and this is genus wanting to aspartate transaminate. So, the top reaction, this corresponds to your aspartate transaminates and when this particular gene is knocked out, then in FBA there is no way for aspartate to be found. Aspartate can be reincumbrated and ammonium by the aspartase reaction. But since this reaction is not reversible, in FBA there's no way to form aspartate and hence biomass production cannot proceed. That's what is shown. So, but in case of TFBA, the same reaction is reversible and we see that it runs in a diverse direction to form aspartate and hence biomass production can keep this. So, there are about five or six more reactions like this which we have identified through this analysis which we will be performing double gene knockouts to see if these are the reactions which are rescuing which are rescuing the cell from cell deck. So, in summary we have, so the thermodynamic model was constrained with suitable constraints for it to give a better reduction of growth rate. The two models, the comparison gave very similar results. But there were some cases of differences where we found that TFBA is breaking much better and in those cases we were able to identify the reactions that are actually running in the reverse direction which we can proceed with in experiment to find out if these two are actually running in the reverse direction. So, my acknowledgement, I would like to thank Dr. Jennifer Rene for hosting me. My mentor is Joshua Hamilton. He is helping me throughout my stay here and the institutions and agencies involved in this season goal can I work out in the rest of this once in a while? Pardon the bad energy going on here. I USSTF.