This document discusses measuring and controlling variability in a core facility environment. It notes that increased variability decreases statistical power to detect real effects. Common sources of variability include biological differences, sample preparation inconsistencies, and technical issues. The document asks how best to communicate these issues to customers given limited funding to directly measure variability. It discusses challenges like analyzing samples over long periods when preparation methods may differ, having no control over initial sample collection, and potential sources of variability like pipetting errors. Finally, it provides an example method using pooled and individual samples to empirically measure variability introduced through sample processing.
1. Controlling or at least
measuring
Variability In a core facility environment
2. Variability
• Increased variability = decreased power
• Power = probability of find an effect that is there
• You can fight this by increasing the sample size but
often it is much cheaper to decrease variability
instead
4. What is possible in a core
facility?
• Almost no one who sends me samples has enough
money to measure variability or wants to pay for it
• What are the best ways to communicate these
issues to customers?
• How do you know what your variability is if there are
no resources to measure it?
• How do you measure variability when you have a
large number of different types experiments?
• How much QC do you bundle into your costs if you
have to charge people
5. Some issues I routinely
have
• Analyzing samples over months at a time….
• Sample preparation of Plant tissue may be
completely different than human cells or Plasma, or
Milk In terms of how consistently you can prepare it
• How do I know how consistently I can prepare a
sample
• Often I have no control over how the sample is
prepared. How do I deal with that?
6. Common ways to decrease variability
during sample prep
• Process all samples on the same day by the same
person
o Person can still get tired or make mistakes…Variability may not be
consistent beginning to end
o May not be possible
• Use Robotics for part of the sample prep
o Many things still cannot be done well by robots
• In gel digestion of proteins is not optimal
• Decrease the things you do to a sample
o Fractionation, precipitation, SPE
• Label proteins or peptides upstream and multiplex
7. Common sources of variability you may not be
thinking about
• Pipetting errors
o Can be vary large for small volumes
• Eppendorf 2 ul = 12% Systemic 6% random error
o Hard to get tight cv’s on your spiked peptides
• Variability due to SPE material lots
o The SPE material you use today may not be the same the next time you
buy it
• Variability due to software
o manual integration
o Normalization
9. Is peptide or protein
fractionation worth it?
• Does the fractionation kill your power?
• Is it better not to fractionate ?
• What is the least variable fractionation method for
proteomics?
• How do you measure the variability your
fractionation causes?
10. Example method to
measure variability
• From Chris Becker (Proteometrics)
Pooled
human serum
1
2
3
4
5
n
Sample aliquots
are processed
Processed samples
are pooled before
analysis and
replicates are run
Processed samples
are run individually
Sample Processing