Jake: would you write out the "five scenarios" for me?

8 views
Skip to first unread message

Brian Teague

unread,
Dec 12, 2016, 7:26:43 PM12/12/16
to sbsc-flow...@googlegroups.com
When you have a moment, if you could commit those to (electronic) paper,
I would really appreciate it. I would like to contribute further to our
defining the scope and direction of this ongoing effort, and based on
today's call those scenarios seem to encapsulate the various choices for
you. I think having them in front of me would be useful in clarifying
my own thinking.

Thanks muchly!
Brian

--
Brian Teague
tea...@mit.edu
Weiss Group, Synthetic Biology Center @ MIT

Jake Beal

unread,
Dec 13, 2016, 5:10:39 AM12/13/16
to sbsc-flow...@googlegroups.com
Hi, Brian, and everyone else as well:

Based on our discussion yesterday, here are my current notes on the driving scenarios, measurand, and target precisions:

Scenarios:
Scenario 1: A person compares two sets of their own cell samples, measured on the same machine on different days. Samples are measured with the same defined ERF laser/filter combination. 
Scenario 2: A person compares two sets of their own cell samples, measured on two different machines at their institution. Samples are measured with the same defined ERF laser/filter combination. 
Scenario 3: A person compares a set of their own cell samples measured on their machine to a different person’s set of samples, which that other person measured on their own machine. Samples are measured with the same defined ERF laser/filter combination. 
Scenario 4: A person compares measurements from two different defined ERF laser/filter combinations.  Measurements are taken from one set of samples on one machine. 
Scenario 5: A person compares two sets of cell samples, at least one of which was measured with a laser/filter combination that does not have a set of defined ERF values.

Critically, note that in no case are we actually interested in comparison of a cell sample to a bead. Instead, we wish to compare two cells samples by comparing their ratios with beads. Thus, what we actually need to rely on is that the ratio between cell samples and beads remains stable across space and time.

Our primary focus is on Scenario 1, 2, and 3, which are “must have” for this effort.
My work has frequently used Scenario 4 successfully in practice, so I suspect we will find that we can support that as well; if the evidence isn’t there, we can defer it.
Some members of the group are likely to continue working on Scenario 5, building off of the data we gain in working on the others, but we will not focus on it until the others are sorted out or until encouraging preliminary successes pop up there.



Measurand:
We will take as our prototypical case a strong, relatively tight distribution of expression levels with expected an log-normal distribution, e.g., strong constitutive expression of GFP driven by J23101 in DH5-alpha E. coli. 
Notes: 
1. There is an argument this might technically be a different distribution (e.g., gamma), but with strong expression such distributions closely approximate on another in any case.
2. Strong expression minimizes the impact of instrument noise and background fluorescence on the distribution.

The measurand we target is the geometric mean (mu_g) of scaled values in a sample gated using some form of gaussian mixture model.
The way to quantify the precision with which we are measuring this measurand is the  geometric standard deviation (sigma_g) of a collection of such geometric means.  
A derived value of interest is sigma_g^4, which approximates the range into which 95% of measurements may be expected to fall.




Target precision:

The target precision for Scenarios 1 & 2 is in the range of 1.2x - 1.5x.
Here we want higher precision, to support combining samples into a single data set, precision modeling and design, etc.
- The 1.2x lower bound is set because the NIST/ISAC study on beads shows it will be hard to get hold of beads that perform any better than this.
- The 1.5x upper bound is set bases on giving about the same amount of error budget to sample-to-sample variation within a well-performed experiment sequence. This is experimentally supported by the 2015 iGEM interlab, which found a 1.24-fold standard deviation across replicates for hundreds of sets of replicates.

The target precision for Scenario 3 is in the looser range of 1.5x - 2.0x
Here we can tolerate lower precision, with a goal of a sanity check that biological constructs being imported from another organization are operating correctly.
- The 1.5x lower bound comes from allocating a similar error budget to that of the beads to the difference in how two different people use their different equipment in different labs, given typical difficulties in protocol reproduction in biology.
- The 2.0x upper bound comes from the fact that at that level, the 95% range is an order of magnitude, and beyond that level the utility of comparison starts to drop significantly.

The target precision for Scenario 4 in the range of 1.2x - 2.0x.
Here we’d like to support precision modeling and design, but might not be able to do as well.
- The 1.2x lower bound comes from the possibility of direct conversion between ERF values on beads that may be well-correlated in their variation between channels
- The 2.0x upper bound comes from the possibility of using biological co-expression controls, which face similar experimental complexity issues to Scenario 3.

No goals are set for Scenario 5 until we have preliminary evidence in hand.

Thanks,
-Jake


--
You received this message because you are subscribed to the Google Groups "SBSC Flow Cytometry Working Group" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sbsc-flow-cytometry+unsubscribe...@googlegroups.com.
To post to this group, send email to sbsc-flow-cytometry@googlegroups.com.
Visit this group at https://groups.google.com/group/sbsc-flow-cytometry.
To view this discussion on the web visit https://groups.google.com/d/msgid/sbsc-flow-cytometry/4df141a1-68fe-066e-cf6e-f9280ab801a0%40mit.edu.
For more options, visit https://groups.google.com/d/optout.

Ross, David J. (Fed)

unread,
Dec 14, 2016, 6:39:57 AM12/14/16
to sbsc-flow...@googlegroups.com

Jake-

Can you put that list of scenarios on the Google Drive for comment/discussion?

 

Thanks-

To unsubscribe from this group and stop receiving emails from it, send an email to sbsc-flow-cytom...@googlegroups.com.
To post to this group, send email to sbsc-flow...@googlegroups.com.

 

--

You received this message because you are subscribed to the Google Groups "SBSC Flow Cytometry Working Group" group.

To unsubscribe from this group and stop receiving emails from it, send an email to sbsc-flow-cytom...@googlegroups.com.
To post to this group, send email to sbsc-flow...@googlegroups.com.

Jake Beal

unread,
Dec 14, 2016, 7:02:30 AM12/14/16
to sbsc-flow...@googlegroups.com
Sure; I've added it with the name "Driving scenarios, measurand, and target precision"

Thanks,
-Jake


To unsubscribe from this group and stop receiving emails from it, send an email to sbsc-flow-cytometry+unsub...@googlegroups.com.
To post to this group, send email to sbsc-flow-cytometry@googlegroups.com.

--
You received this message because you are subscribed to the Google Groups "SBSC Flow Cytometry Working Group" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sbsc-flow-cytometry+unsub...@googlegroups.com.
To post to this group, send email to sbsc-flow-cytometry@googlegroups.com.

--
You received this message because you are subscribed to the Google Groups "SBSC Flow Cytometry Working Group" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sbsc-flow-cytometry+unsub...@googlegroups.com.
To post to this group, send email to sbsc-flow-cytometry@googlegroups.com.

Visit this group at https://groups.google.com/group/sbsc-flow-cytometry.
To view this discussion on the web visit https://groups.google.com/d/msgid/sbsc-flow-cytometry/BY1PR09MB04534F272E04CD0C5BF20AF6E79A0%40BY1PR09MB0453.namprd09.prod.outlook.com.
Reply all
Reply to author
Forward
0 new messages