Engineering Design Process Introduction
At the UFlorida this year, we decided to gather insight into the nature of sepsis and it's affects on the immune system by building an in vitro organoid system, and constructing a system of ordinary differential equations to go with it. These ODE were built using the Engineering Design Process Cycle. On this page we will describe exactly how we went through iterations of designing, building, testing, and learning in order to create a set of ODE that describes the dynamical process of both the physiological and pathophysiological states of the immune system. By treating the immune system as a set of nodes that each have a positive or negative “force” on the other nodes, we were able to build a mechanistic model based on literature that allowed us to simplify the complex immune system into something that could more readily be described using differential equations. The reason our model is different is because it takes the number of HSPC’s into account along with pro-inflammatory leukocytes and anti-inflammatory leukocytes. Having the HSPC’s in our model has given us insight into the nature of how the level of stem cells in our immune system can affect the overall immune response.
Our engineering design process was split into two sections. One iteration of the process was used to build our first set of equations. The reason we required two separate design procceses was because once we could generate our first set of graphs from our equations, it was much easier to analyze how the ODE were behaving. From here we could begin our next engineering design process, using the output graphs in our testing phase.
Design 1
The first step of our our design process was creating a mechanistic model of the immune system that we would base our ODE on. This mechanistic model was created using previous literature with the goal of describing the immune system in a simple enough way that we could create simple enough equations for it, while still being able to detect realistic trends in the data. After creating this mechanistic model, we then had a plan to start building our ODE. After we were done discussing our discrepencies in the ODE and the literature, we would decide how these could potentially be fixed before going back into the build phase.
Build 1
In this step, we would each dig through previous literature and put together our ODE using what we found. We would guide our search using what we decided during the design step.
Test 1
After each iteration of this first engineering design cycle, we would meet during the test phase after we had all independently worked during the build phase. At these meetings we would compare notes that we had taken, and discuss discrepancies between certain papers, our mechanistic model, and our equations.
Learn 1
After putting together what we had found, our discussions allowed us to find what needed to be changed, and begin the design process again.
Engineering Design Process Two
This iteration was dedicated to tweaking our ODE to reflect what we believed should be found in a clinical setting and was fueled more by the graphs we were outputting. We iterated through this process 5 times over 5 weeks.
Iteration 1
Test, Learn, Design 1
In the first iteration, we focused on a few things. First of all, we noticed that the graphs did not make sense. We appreciated that the pathogen curve presented like we hoped for a death scenario, but the HSPC graph and the cytokine graphs did not line up. Since the HSPC graph fell to zero, we believed the others should also go to zero. We focused in on a couple potential problems with the model. First, we were unsure if the code implementation of the ODE was entirely correct, this needed to be checked again. Second, we needed to check the decay terms for the cytokines. Our plan was to fix these things and attempt to get a steady-state graph (with no pathogen insult).
Build 1
Over the course of the next week, we would build the changes discussed previously on our own.
Test, Learn, Design 2
In the second iteration, we came closer to what we thought were proper outcomes. We ran into the problem that our anti-inflammatory cytokines were not decreasing, even when it's decay term was increasing, but we believed this run had promise as it was the first time we "killed" a pathogen insult and our HSPC level was perturbed but returned back to homeostasis. We needed to fix the cytokines not returning to homeostasis though. Also, while not shown, our Leukocyte graph showed an increase in leukocytes after the infection was killed and the HSPC level returned to normal, which did not look corect. We had also fixed the problem from before of the inflammatory cytokines not decreasing after the HSPC population had died off. This can be seen in the second set of graphs.
Our first time seeing a return to homeostasis in the HSPC population
First time seeing extinction of all immune system components upon a major pathogen insult
Build 2
Over the course of the next week, we would build the changes discussed previously on our own.
Test, Learn, Design 3
In the third iteration we believed was the first time we achieved many accurate cases. It was our first time achieving a steady-state without pathogen insult. We saw our predicted trend of oscillating levels of certain cytokines and leukocytes. We had a more accurate depiction of a moderate pathogen insult, as the inflammatory levels returned to normal faster, and also showed our predicted oscillating behavior. We also had our first evidence of sepsis chronic death, and sepsis chronic (no death or recovery in the time scale used). These were important devlopments, as our model was starting to achieve steady-states for multiple cases that we predicted it should. From here, we planned to try to achieve more of our predicted cases by tuning initial values, now that we had built confidence that our model had achieved many of our other predicted cases.
We were able to produce three predicted outcomes in the third round of graphs. One for a healthy person with no pathogen insult, one for someone with chronic sepsis, and one for someone with chronic sepsis with a second pathogen insult. These are presented below respectively.
Healthy person with no pathogen insult:
Chronic Sepsis:
Chronic Sepsis with second infection:
