6.28-7.20
Task:
1. Generate a mutation library for a given sequence.
I used a traverse algorithm to enumerate every possible sequence with different mutations in a 7-nt region near the prospective RBS. For a given sequence, 47(16384) mutant sequences are produced.
2. Determine △Gtot for a given mutant.
I first tried to simplify the model produced in the work by Voigt et al(2009). According to the free energy model brought forward by Voigt et al(2009), the △G required for translation initiation consists of five major terms:
△GmRNA(the folding free energy change for a free mRNA sequence in solution)
△GmRNA:rRNA(the energy change brought by the binding of the 16S rRNA tail to the specific binding region in the 5’ UTR of the mRNA)
△Gspacing(the free energy change brought by the spacing sequence between the RBS and the start codon, determined by the sequence length using an empirical formula derived from experimental data)
△Gstandby(the energy needed to expose the 16S-rRNA binding site to the 16S-rRNA. However, according to calculations by Voigt et al, this term is almost always approximately zero, therefore it is omitted in our model)
△Gstart, which is the free energy change from binding of the 16S-rRNA to the start codon(it is treated as a constant for all mRNAs with AUG as the start codon)
In my own calculations, the UNAfold package is used. hybrid-ss-min.exe calculates △GmRNA, and UNAfold.pl calculates △GmRNA:rRNA. With the help of a friend proficient in UNIX and programming, I was able to determine the optimal combination of △GmRNA:rRNA and △Gspacing to minimize their sum, and run the programs in a batch for all the above terms to determine △Gtot for every sequence in the library.
However, the results are unsatisfactory as they are significantly different from the data produced in Voigt’s work. After close examination of the software package provided by Voigt et al, I found that other than over-simplification in our model, the model described in the paper has significant deviations from that used in the software’s algorithms. Besides, the software employs NuPACK instead of UNAFold for the energy calculations.
Therefore, we decided to use the provided software package to generate data for our own sequences. I had to tackle problems including properly running the software package and processing our sequences in a batch before I finally obtained △G data for experimental verification. Great acknowledgement must be paid to Qiu Yunjiang for the help given in programming and data processing.
07.21
Constructed and repeated the computational modeling results of the AND Gate module in iGEM 2009 Team Peking’s Project, using Matlab to solve the ordinary differential equations(ODEs).
Simplified the sequential logic circuit containing a bistable switch module constructed by iGEM 2007 Team Peking, modeling the bistable switch to analysis the effects of translation initiation rates on the switch’s properties.
07.25
Constructed a coarse phase diagram of the bistable module to reflect ratios of red and green fluorescence, thus indirectly reflecting the strength of the ribosome binding site(RBS).
07.31
Platform for high-throughput Gibbs free energy calculation to predict translation initiation complete. First batch of sequences containing computationally mutated RBSs are entered into the software package provided by C. Voigt et al. and calculated for gene expression strengths.
08.03-08.09
Running the prediction platform on several batches of sequence libraries to obtain sufficient amounts of data for further experimental validation of our fine-tuning knob.
08.14
Rewrote the first draft for Human Practice page, integrating information collected from other teammates and summarizing everyone’s views on biosafety issues in synthetic biology.
Improved the structure and design of the Human Practice page so that it could convey more information in a more efficient way.
08.15
Collected data for the survey conducted by Xiao Qingyang and other teammates and presented the statistical data, drawing conclusions and providing suggestions for ensuring biosafety in laboratories using antibiotic-resistant bacteria.
08.18
Modeling of the AND Gate module reveals critical parameters that establish mapping between RBS strength(△G) and fold-changes in output when input turns from 0 0 to 1 1. This was accomplished by quantitatively studying the “transfer function” derived in previous work by Voigt et al..
08.19
Fitted the calculated RBS strengths to experimental measurements of actual expression levels(using green fluorescent protein, GFP). Results were partially disappointing as the two sets of data showed much less correlation than expected. The calculation and fitting were repeated changing sequence cutoff values(length of sequence chosen for calculation of △G, but no improvements were observed. Planned to conduct experiment repeats to verify the reliability of measurements. However, one of the calculated expression level data sets showed excellent linear correlation with experimental values, with Pearson coefficient Square R≈0.9
08.20
After double-checking with Robin on previous sequence data, a mistake in programming was found that altered a few nucleotides in the protein coding sequence of GFP. The mistake resulted in unexpectedly low correlation in some sequence libraries but high in some others. After correcting the mistake, most sequence libraries showed fairly good correlation.
08.22
Together with Robin, we tried to further validate the prediction of gene expression levels by fitting parameters β in the empirical equation: Expression level∝exp{-β△G}. Theoretical value of β is around 0.4, and fitted (to experimental measurements) values were between 0.2 and 0.4, implying that some corrections to the prediction method need to be made.
08.24-08.27
Calculated libraries for RBSs with lengths from 1nt~7nt, with a total of over 30000 rounds of △G calculations.
This has consumed most of my laptop’s computation powers those days!
08.29
Modelled the theoretical fold-change in output of the AND gate module when input turns from 0 0 to 1 1. Results showed that as the RBS of the T7ptag coding sequence becomes stronger, the circuit displays decreasing levels of AND gate property, in accordance with both our expectation and experiment results.
09.03
After summarizing previous calculations of sequence libraries, we decided on a cutoff value of 51nt, that is, only calculating the sequence 51nt upstream and downstream of the start codon, which provided satisfactory fitting with experiment measurements for most sequence libraries. Thermodynamic paramters of RNA1995 was also employed to improve accuracy of prediction.
09.05-09.12
New semester starts. Busy with school affairs.
09.14
Plotted the colormap of the AND Gate phase diagram with input levels as horizontal and vertical axes. Parameters from work by Voigt et al. were used. The procedure was repeated on RBSs with different strengths. This was accomplished by replacing original parameters in the transfer function with new ones that proportionally changes with △G values of the RBSs. The results qualitatively described the behaviour of the AND gate module as RBS strength changes, but not in good accordance with experiment results using the module constructed by ourselves. Further experiments that measure input levels more precisely are needed.
09.17-09.20
Establish a quantitative correlation between fold-change of output upon induction of the AND gate module and △G values of the RBS, thus enabling the selection of appropriate RBSs for the circuit to perform particular output functions as determined by dosage response to ribozyme ligands.
9.23
Planned the Wiki outline, including documentation of the RBS Calculator platform and the mathematical modelling for the AND gate and bistable switch modules.
9.25
Discussed modeling with Professor Ouyang during JC. The bistable switch should be verified using a stochastic algorithm instead of a steady-state one.
9.26~9.28
Optimized details to the AND gate modeling using new experimental data: quantification of the AND gate experimental results by employing the fold-changes in output under different inducer combinations. The fold-changes were plotted versus parameter a to show its correlation with translation rates.
10.1~10.4
Wrote first and second drafts of the modeling part of the wiki. After revising the text, figures and images were sequentially designed and added to the text. The wiki draft was reformatted into an academic style and Appendices were written for more details of our models for the AND gate and bistable switch, ready for the final round of revision. |
"
