Sedimentation Velocity Analysis Flowchart
1. Import Experimental Data
2. Edit Data
3. Time Derivative Analysis
This step will provide an range for the sedimentation coefficient for use in the upcoming analysis steps.
-
Open Velocity: Time Derivative and select Load Experiment to load the experimental data.
-
Adjust analysis controls:
Set Data Smoothing to ~ 10.
Set Boundary Pos (%) to 0.
Create an Exclusion Profile for scans that do not exhibit a stable upper plateau.
Select Average S to plot the dc/dt S-value distribution (default setting).
Adjust S-value cutoff as needed.
The correct S limits to choose are the left and right limits of the S-value distribution where all signal returns to baseline. The minimum S-value allowed is 0.2 S.
4. 2DSA: Fit Time Invariant Noise
This step can be done locally on UltraScan's 2DSA module, but the following assumes the use of the USLIMS site.
-
Log into the US3 LIMS portal and navigate to Analysis: Queue Setup.
-
Load the Dataset.
-
Setup 2DSA Control
a. Set the \(s\)-value Limits from the range obtained in Step 3.
b. Set \(f/f_{0}\) Limits to 1-4 or adjust the upper limit based on prior knowledge of the sample.
c. Set the Resolution. The default for both \(s\) and \(f/f_{0}\) is 64. The resolution is the number of points into which this variable will be discretized. For example, if the range is selected from 1-5, and the resolution is set to 40, there will be 10 grid points/\(s\)-value, resulting in an increment of 0.1 S.
d. Set the Grid Resolution. The default resolution of 64 will be sufficient for most situations. When you have a very poly-disperse or heterogeneous sample you may have to go higher. Larger values will increase compute time. A good strategy is to check if there are noticeable jumps in the fit. If there are, increase the resolution. If you are trying to fit a bi- or tri-modal system, you may be able to get a higher resolution by using the custom grid method, and use the appropriate resolution for each grid.
e. Select Fit Time Invariant Noise.
f. Leave all other settings at default values.
-
Submit the Job to your desired cluster. Check Queue Viewer for job completion.
-
Open Velocity: FE Model Viewer to confirm the results. Load the Dataset of the newly generated time invariant noise file. The file will be of the format "2DSA.run_name". Remember to Save Results.
If random residuals are obtained, proceed with step 5, but if residuals do not look random, and a strong diagonal line in the residual bitmap is apparent, investigate range settings for S and f/f0 settings and repeat 2DSA with improved ranges.
Note: Do not set the lower s-value limit too low, as this could create artificially low-s species signal if a baseline or slowly changing baseline exists. These artifacts are better handled in the time invariant noise.
5. 2DSA: Fit Meniscus, TI Noise, RI Noise
-
Log into the US3 LIMS portal and navigate to Analysis: Queue Setup to Load the Dataset. By default, it will load the dataset that was saved from the previous step.
-
Setup 2DSA Control
a. Set the \(s\)-value and \(f/f_{0}\) Limits. Use the same range settings as in Step 4.
b. Set the Resolution and Set the Grid Resolution.
c. Select Fit Time Invariant Noise and Fit Radially Invariant Noise,
d. Select Fit Meniscus. The fitting should occur over 0.03 cm with 10 points.
-
Submit the Job to your desired cluster. Check Queue Viewer for job completion.
-
Click on Scan Database (if using the database) and check the status line for new results.
-
Go to File: Load to load the desired meniscus fit.
-
After updating the meniscus, confirm the deletion of the scans that resulted in non-optimal RMSD values.
-
Open Velocity: FE Model Viewer to confirm the results. Load the Dataset by selecting the file of the format "2DSA.FM.run_name". Remember to Save Results.
6. 2DSA: Iterative
-
Log into the US3 LIMS portal and navigate to Analysis: Queue Setup to Load the Dataset.
-
Setup 2DSA Control
a. Set the \(s\)-value and \(f/f_{0}\) Limits. Use the same range settings as in Step 5.
b. Set the Resolution and Set the Grid Resolution.
c. Select Fit Time Invariant Noise and Fit Radially Invariant Noise,
d. Select Iterative Refinement and Set Refinement Level to 10 iterations.
-
Submit the Job to your desired cluster. Check Queue Viewer for job completion.
-
Open Velocity: FE Model Viewerto confirm the results. Load the Dataset by selecting the file of the format "2DSA.IT.run_name". Remember to Save Results.
All subsequent analyses methods should now be based on the model generated in this final 2DSA refinement step.
At this point, multiple analysis options exist depending on the properties of the analyte distribution. If a paucidisperse solution is obtained, parsimonious regularization with the genetic algorithm method is appropriate. Otherwise, the data should be analyzed only by the 2DSA analysis in conjunction with a 50-iteration Monte Carlo analysis. Both options are explained below.
7. Genetic Algorithm Analysis (optional)
Perform this step if the refined 2DSA data is appropriate for genetic algorithm analysis.
-
Load the Model from Step 6 into the initialization program.
-
Assign Initialization and Save Results.
-
Log into the US3 LIMS portal and navigate to Analysis: Queue Setup to Load the Dataset to Genetic Algorithm analysis.
-
Select and Load the gadistro File from the UltraScan/results/run-id directory for the correct triple.
-
Open Velocity: FE Model Viewer to confirm the results. Load the Dataset by selecting the file of the format "2DSA.GA.run_name". Remember to Save Results.
8. 2DSA Monte Carlo Analysis (recommended)
-
Log into the US3 LIMS portal and navigate to Analysis: Queue Setup to Load the Dataset.
-
Select 50 Monte Carlo Iterations.
If the 2DSA distribution appears to be a sparse solute situation, and not a smooth continuous distribution of many species, you can further refine the data with a parsimonious regularization using the GA analysis.
When using 2DSA Monte Carlo distributions for the GA initialization, make sure to use the manual GA initialization method in Velocity: Initialize Genetic Algorithm.
-
Select and Load the gadistro File from the UltraScan/results/run-id directory for the correct triple.
-
Open <span style="color: #00008B";>Velocity: FE Model Viewer to confirm the results. Load the Dataset by selecting the file of the format "2DSA.MC.run_name". Remember to Save Results.
9. Perform Monte Carlo GA Analysis (optional)
-
Using the results from Step 7, Initialize Genetic Algorithm - Monte Carlo.
-
Log into the US3 LIMS portal and navigate to Analysis: Queue Setup to Load the Dataset from Step 6 (default).
-
Select a multiple of 8 for the Monte Carlo iterations (48, 56 or 64 are good choices).
-
Select parallel processing with 8 program groups.
-
Select and Load the gadistro File from the UltraScan/results/run-id directory for the correct triple.
-
Submit to desired cluster.
-
Open Velocity: FE Model Viewer to confirm the results. Load the Dataset by selecting the file of the format "2DSA.GA.run_name" generated from the GA distribution model. Remember to Save Results.
10. Perform van Holde-Weischet Analysis
-
Load the desired experiment, applying the noise files from Step 6.
-
Check the boxes for Plateaus from 2DSA and Use Enhanced vHW
-
Adjust analysis controls:
Back Diffusion Tolerance to between 0.101 or 0.201
Increase Divisions to 150
Increase Data Smoothing to 11
% of Boundary
Boundary Position
If appropriate, delete early scans to improve resolution and reduce noise. Only keep scans and boundary portions that contribute to well correlated line fits in the linear extrapolations.
-
Select Groups, if appropriate, to generate weight averaged s-values for discrete species
-
Display Distribution Plot.
A mono-disperse solution is shown by a straight line on the distribution plot. Ensure that the line is smooth without any kinks or back-bends. If there are kinks or back-bends, adjust Divisions, Data Smoothing, and Scan Exclusion profiles.
-
Select Display Histogram in the Distribution Plot module.
Adjust Histogram Sensitivity to ~40
-
Save Data and distributions.
11. Overlay Combined Distributions
All van Holde - Weischet distributions and finite element models can be combined into a single plot for easy comparison.
-
Open Velocity: Combine Distribution Plots (vHW) for van Holde - Weischet plots.
-
Use Velocity: Combine Discrete Distributions for all finite element models (2DSA, GA, Monte Carlo).