1. Bottom-Up Proteomics Data collection
Ruedi Aebersold, Ph.D
Institute of Molecular Systems Biology, ETH-Zürich;
Faculty of Science, University of Zürich

2. The proteome: The ensemble of all biochemical reactions

3. Steps of Bottom-up proteomics
protein sample
protein identifications
Database
Protein
level A B C D A B C
Peptide grouping/
validation
enzymatic
digestion
Peptide
level
Quantitation
Validation
Database search
peptide mixture
peptide identifications
LC/MS/MS
MS/MS spectrum
level
MS/MS spectra
Protein Inference assumptions?? -- many possible, none is right

4. The proteome as seen by a mass spectrometer:
Possibly:10exp6- 10exp8 features
1200
m/z
1100
1000
900
800
700
600
500
400
min 10 20 30 40 50 60 70 80 90
100
110

5. Slicing and dicing the proteome

6. Proteomics: The global (quantitative) analysis
of the proteins expressed in a cell at a time
Enumerate all the
components
of a proteome
- Proteome as
Database
-Analytic chemistry slant
Proteome analyzed
once
Detect dynamic
changes in proteome
following external or
internal perturbations
- Proteomics as
Biol. or clin. Assay
- Biology slant
multiple (infinite) times
Haynes P, Gygi S, Figeys D, and Aebersold R. (1998) Proteome analysis: biological assay or data archive? Electrophoresis 19:

7. Proteomics: The global (quantitative) analysis
of the proteins expressed in a cell at a time
Enumerate all the
components
of a proteome
Proteome as
database:
Proteome analyzed
once
Detect dynamic
changes in proteome
following external or
internal perturbations
Proteomics as
Biol. or clin. assay:
multiple (infinite) times

8. Human PeptideAtlas 2013-2015 14,274 13,230 +377 +110 +4 +34 +3 +516
2014
2015
2013
2014
2015
2013
2013

9. Human PeptideAtlas
14,274
13,230
+377
+110 +4
+34 +3
+516
True new identifications or statistical noise?
2013
2014
2015
2013
2014
2015
2013
2013

10. Open questions re: Proteome catalogue…
When have we reached an endpoint in proteome cataloguing?
Why do we reach apparent saturation before hitting all predicted ORF’s?
What are relevant endpoints? (one representative per ORF?, all proteoforms? Other?)
How do we quantify proteins?
How do errors propagate in large datasets and how do we control FDR at peptide and protein level?
How do we best complete the catalogue?
What (biology) can we learn from the (complete) catalogue?

11. Proteomics: The global (quantitative) analysis
of the proteins expressed in a cell at a time
Enumerate all the
components
of a proteome
Proteome as
database:
Proteome analyzed
once
Detect dynamic
changes in proteome
following external or
internal perturbations
Proteomics as
Biol. or clin. assay:
multiple (infinite) times

12. Data and the scientific method
Scientists are trained to recognize that correlation is not causation, that no conclusions should be drawn simply on the basis of correlation between X and Y (it could just be a coincidence). Instead, you must understand the underlying mechanisms that connect the two. Once you have a model, you can connect the data sets with confidence. Data without a model is just noise.
But faced with massive data, this approach to science — hypothesize, model, test — is becoming obsolete.
There is now a better way. Petabytes allow us to say: "Correlation is enough." We can stop looking for models. We can analyze the data without hypotheses about what it might show. We can throw the numbers into the biggest computing clusters the world has ever seen and let statistical algorithms find patterns where science cannot.

13. Data matrix supporting analyses via association
Accurately and reproducibly quantify proteotypes across samples and conditions
Conditions:
Clinical cohorts
Time courses
Dosage courses
Samples with structured genomes
Conditions 1-n
Proteins 1--n

14. Open questions re: Association studies
How many proteins are enough?
Which ones?
How precisely do proteins need to be quantified?
Which peptides are best suited to quantify a protein?
Should proteins be considered as independent actors (like transcripts) or as parts of modules?
What factors affect protein modules and how?
How do errors propagate in large datasets and how do we control FDR at peptide and protein level?
Are data reliable, robust and accessible enough?
Data integration, dissemination of methods.