Brianna N. Samuels-Week 7
to analyze the models and come up with a tweak for the project thus far
Materials and Methods
Analyzing Results of First Model Run (Due for the Week 7 deadline, midnight March 7
Here is what you need to consider when analyzing the results of your model.
- What is the overall least squares error (LSE) for your model?
- You will find this on the "optimization_diagnostics" worksheet of your output workbook.
- Since the input data are noisy, the model can only minimize the error so far. It is more "fair" to look at the ratio of the least squares error to the minimum theoretical least squares error that the model could have achieved given the data. We call this the LSE:minLSE ratio. You should be able to compute it with the values given on the "optimization_diagnostics" worksheet.
- We will compare the LSE:minLSE ratios for the ten models run by everyone in the class.
- You need to look at the individual fits for each of the genes in your model. Which genes are modeled well? Which genes are not modeled well?
- Look at the individual expression plots to see if the line that represents the simulated model data is a good fit to the individual data points.
- Upload your output Excel spreadsheet to GRNsight. Use the dropdown menu on the left to choose the data you will display on the nodes (boxes). Compare the actual data for a strain with the simulated data from the same strain. If the model fits the data well, the color heatmap superimposed on the node will match top and bottom. If the fit is less good, the colors will not match.
- What explains the goodness of fit to the model?
- How many arrows are incoming to the node?
- What is the ANOVA Benjamini & Hochberg corrected p value for the gene?
- Is the gene changing its expression a lot or is the log2 fold change mostly near zero?
- Make bar charts for the b and P parameters.
- Is there something about these parameters that explains the goodness of fit for the individual genes?
Tweaking the Model and Analyzing the Results
For the Week 7 deadline (midnight, March 7) state which of these "tweaks" you would like to try and explain why. You don't have to have done it yet, but you need to pick one and explain why you chose it. It is a good idea to choose the same "tweak" as a group so that you can help each other compare results.
You will carry out an additional in silico experiment with your model. Let Dr. Dahlquist or Dr. Fitzpatrick know what you are planning to do to get approval and suggestions on how to do it. You will report out your results in a research presentation in Week 9. Some ideas are:
- For our initial runs, we estimated all three parameters w, P, and b.
- How do the modeling results change if P is instead fixed and w and b are estimated?
- How do the modeling results change if b is fixed and w and P are estimated?
- How do the modeling results change if P and b are fixed, and only w is estimated?
- For our initial runs, we included all three microarray datasets, wt, Δgln3, and Δhap4.
- What happens to the results if we base the estimation on just two strains (wt + one deletion strain)?
- What happens to the results if we base the estimation on just the wt strain data?
- When viewing the modeling results in GRNsight, you may determine that one or more genes in the network does not appear to be doing much.
- What happens to the modeling results if you delete this gene from the network and re-run the model (remember you will have to delete references to this gene in all worksheets of the input file).
- You also might think that a particular edge (regulatory relationship) is not needed. What happens if you delete that edge?
- What happens if you include the t90 and t120 expression data?
LSE/minLSE values and ratio
- LSE = 0.641542
- minLSE = 0.475506
- LSE:minLSE ratio = 1.3
GRNsight Heatmap Analysis
- Had one arrow pointing into it (MSN2). The arrow is red suggesting that MSN2 activates ACE2. The arrow is moderately thin/thick. The heat map suggests that it wasn't that great a fit for the wild type. It was all blue but different shades and the top didn't match the bottom for the wild type. For the dGLN3, it didn't match at all suggesting it was a bad fit. For the HAP4, some matched suggesting it was a better fit compared to the GLN3 and wild type. The ZAP1 was an ok fit in the sense that some matched but not much just like the wild type
- Had one arrow stopping at it (MSN2). The arrow was blue suggesting that MSN2 represses ASG1. The arrow is thin. The heat map suggests that it was a bad fit for the wild type, a decent match for the GLN3, and a not so good match for HAP4 and ZAP1
- Had 2 arrows stopping at it from MSN2 and STB5 and was pointing an arrow to MSN2. The arrow coming from MSN2 and STB5 were both blue suggesting they were repressing GLN3. The MSN2 arrow was very thick while the STB5 arrow was thin. The arrow pointing to MSN2 was very thin and grey suggesting that it's unknown what exactly it does but it was too small to determine. The wild type and GLN3 are bad matches, while the HAP4 was not a good match. ZAP1 had an okay match in the sense that they mostly matched and the color shades were somewhat different.
- Had 2 arrows pointing to it. One thick blue arrow from YAP1 and one thick red arrow from STB5. This suggests that YAP1 represses HAP4 and STB5 activated HAP4. Suggests a not so good fit from the wild type, a decent fit from GLN3 and ZAP1, and a really bad fit for HAP4.
- Had 2 arrows pointing to it and 10 arrows pointing to others. The 2 arrows pointing to it were from GLN3 and YHP1 which were both thin grey lines suggesting that it's unclear whether it activates or represses because it was so little. The other arrows point to ACE 2 (a moderately thick red arrow), YOX1 (a very thick red arrow), YLR278C (a very thin red arrow), UME6 ( a very thin red arrow), YHP1 (a thick red arrow), SWI5 (a thin red arrow), SFP1 (a thin blue arrow), GLN3 (a slightly thick blue arrow), YAP1 (a thick blue arrow), and ASG1 (a slightly thick blue arrow). The wild type, GLN3, HAP4, and ZAP1 were not good matches
- Had one arrow pointing to it and was pointing one arrow to another. The arrow it was pointing to was YAP1 and it was a thin red arrow (suggesting it activates YAP1). The arrow pointing to it was from ZAP1 and it was a thin grey arrow (suggesting that it's uncertain if it represses or activates PDR3). The wild type, HAP4, and ZAP1 were not a good match but the GLN3 was an okay match
- Had 1 arrow pointing to it and 2 pointing to others. The arrow pointing to it was from MSN2 and it was a thick blue one suggesting it is repressed by MSN2. The arrows pointing away from it were to TUP1 and YHP1. Both were blue suggesting it was being repressed by SFP1. The YHP1 was very thick and the TUP1 was slightly thick. The wild type was a decent fit, the GLN3 and HAP4 were not so good a match, the ZAP1 was a bad match,
- Had 0 arrows pointing to it but had 4 arrows pointing away from it. There was a very thick red arrow pointing to HAP4, a very thin blue one to GLN3, and two slightly thick blue ones to UME6 and YOX1. ZAP1 and GLN3 were a bad fit, HAP4 was a not so good fit, and the wild type was an okay fit
- Had 1 arrow pointing to it from MSN2. The arrow was slightly thin and red suggesting it is activated by MSN2. The wild type,HAP4 and ZAP1 were a not so good fit, the GLN3 was a pretty good fit.
- Had 2 arrows pointing to it and 1 pointing to others. The arrow pointing to others was from YLR278C which was a thin red arrow suggesting it is activated by TUP1. The other arrows are pointing from SFP1 which was a thick blue arrow suggesting repression and from YAP1 which was a thin red arrow suggesting it is activated by YAP1. ZAP1 was a pretty good match, wild type, HAP4 and GLN3 was a not so good match
- Had 4 arrows pointing to it. 2 thin red arrows from ACE2 and MSN2 suggesting they activate UME6. 2 blue arrows, a thin one from YAP1 and a thick one from STB5, both suggesting repression. GLN3 was a decent fit, wild type, HAP4 and ZAP1 were not so good a fit
- Had 3 arrows pointing to it and 4 arrows pointing away. PDR3 activates YAP1, MSN2 represses YAP1, YAP1 represses itself but very minimal (hence the grey arrow), YAP1 represses YOX1, UME6, and HAP4 and activates TUP1. The wild type, GLN3, HAP4, ZAP1 seem to be a pretty good fit
- Had 2 arrows pointing to it from MSN2 (which activates it) and SFP1 (which represses it)and 1 arrow pointing to MSN2 (which is grey). The wild type and GLN3 have okay fits. HAP4 and ZAP1 have not so good fits.
- Had 1 arrow pointing to it from MSN2 which was a thin red arrow showing it activates YLR278C. HAP4 and ZAP1 are not so good fits while the wild type and GLN3 had an okay fit.
- Had 3 arrows pointing to it from MSN2 (which activates it and had a very thick arrow), STB5 (which represses it) and a thick one from YAP1 (which represses it). Wild type, ZAP1, HAP4 and GLN3 had okay fit
- Had 1 arrow pointing to it from itself (which was a thin red one implying it activates itself) and another arrow pointing to PDR3 (which was grey but very small activation). Wild type, GLN3, HAP4 has okay fit while ZAP1 has a pretty good fit
- I would say that the gene expression is changing its expression a bit
- I would say that there doesn't seem to be a correlation between the fitness of the model and the corresponding parameters.
Graphs and Tables
- Our group decided to tweak our project by deleting that ZAP1 strain from our data. We chose this because for some of us we noticed that it wasn't really having a big impact on the results in the sense that it wasn't activating or repressing as many genes as some of the others. We formed a hypothesis and we believe that deleting the ZAP1 data will not affect the results that we received when including its data. We wanted to observe whether our hypothesis was true or not. To do this, we have to go through all the excel sheets and delete all information pertaining to ZAP1 including the individual sheet created for it. Once we do that, we will re-run the excel sheet into matlab and compare the results to our previous run. We aren't expecting much of a change
The purpose of this assignment was to make models and analyze the models. From this, we were able to receive actual values instead of estimated guesses for the weights, production rates, and thresholds. We were able to get a better understanding of which genes were activators and which were repressors as well as which were activants and repressants. We attempted to understand the models and will continue to analyze the data. Overall, I would say we accomplished our goal. In the future, we hope to see the effects of deleting the ZAP1 gene to get more optimized results.
- Homework Partners: Desiree Gonzalez and Ava Lekander We texted each other to determine how to tweak our project and be on the same page. We also discussed how to interpret whether things were a good fit or a bad fit.
- Except for what is noted above, this individual journal entry was completed by me and not copied from another source.
- Dahlquist, K. & Fitpatrick, B. (2019). "BIOL388/S19: Week 7" Biomathematical Modeling, Loyola Marymount University. Accessed from:Week 7 Assignment Page
- Assignment Pages:
- Assignment Link:
- Class Journal Page:
- BIOL388 Home Page: BIOL388
- User page: Brianna N. Samuels