Practical Guide to Significantly Improve Peptide Identification

Sensitivity and Accuracy

Bin Ma, CTOBioinformatics Solutions Inc.

June 5, 2011.

The Sensitivity and Accuracy Dilemma

score

false

true

FDR# reported false hits

# reported hits

Publication Guideline• Earlier experiments paid too much attention on sensitivity and

not enough on accuracy.• MCP started the guideline in 2004 to ensure accuracy.

People are generally over-optimistic about how reliable their results are.– ABRF iPRG 2011.

1%

iPRG/ABRF 2011 Study

30 out of 45 submissions have FDR much higher than the required 1%

Estimated FDR lower bound

Estimated FDR upper bound

“ ”

PEAKS Achieved both Sensitivity and Accuracy

1%

PEAKS PEAKS

More peptides in submission

Outline

1. FDR – pitfalls and solutions2. De novo sequencing assisted database search3. Three essential examinations to ensure result

quality.

1. FDR – pitfalls and solutions

FDR Estimation

Search Engine

𝐹𝐷𝑅=¿𝑑𝑒𝑐𝑜𝑦¿ 𝑡𝑎𝑟𝑔𝑒𝑡

target

decoy # decoy hits

Protein DB

Identified Peptides

# false target hits ≈

Pitfall 1 – Multiple Round Search

Round 1. Fast Search

Round 2. More Sensitive Search

FDR underestimation.

# decoy hits# false target hits ¿

more targets than decoys

Craig and Beavis 2004. Bioinformatics 20, 1466–67.

Bern and Kil 2011, J Proteome Res. 10, 2123-27.

Evertt et al. 2010. J Proteome Res. 9, 700-707.

Our Solution: Decoy Fusion

Fast Search

More Sensitive Search

Decoy sequence append to each target protein.

PEAKS DB paper. Submitted.

Equal targets and decoys

# decoy hits# false target hits ≈

Pitfall 2 – Mix Protein and Peptide ID

Idea: Peptides on a multi-hit protein get a bonus on their scores to increase sensitivity.

Pitfall

More multi-hit proteins from target DB more false hits are “saved” from target DBFDR underestimation.

A weak hit is “saved” due to the bonus.

So is this weak false hit.

decoy hit

target false hit

Our Solution: Decoy FusionWeak false hits are “saved” with approx. equal probabilities in target and decoy.

Get the sensitivity, but still estimate the FDR correctly.

Pitfall 3 – Machine Learning with Decoy

Idea: Re-train the coefficients of scoring function for every search after knowing the decoy hits.Pitfall: Risk of over-fit. Machine learning experts only.

Adjust scoring function to remove decoy hits after search.

Fewer target false hits are removedFDR underestimation

Search

target false hits

decoy hits

Solutions

1. Don’t use it. Judges cannot be players.

2. Only use for very large dataset.3. Train coefficients and reuse; don’t re-train

for every search.

oror

PEAKS 5.3

• PEAKS DB used all these techniques (and many more) to ensure the accuracy while maximizing sensitivity.

• Reliable FDR estimation is the top priority in PEAKS DB design.

2. De novo sequencing assisted database search

An Idea to Improve Score Function

score

false

true

Idea: If de novo matches a DB peptide, it is likely to be correct.

De Novo Assisted DB Search# matched amino acidsbetween de novo & DB search

x+4ybest separation line

DB Search Score

score

false

true

Including de novo matching as a feature gives the score function a better discriminative power.

before after

This is just one example of many other new features in PEAKS 5.3 for improving score function.

… far better than what I could ever squeeze out of my data – Stefano Gotta, Siena Biotech

0 500 1000 1500 2000 2500 3000 3500 40000.0%

0.5%

1.0%

1.5%

2.0%

2.5%

# of PSM

FDR

product M PEAKS DB

“ ”

DB search

Found?

Yes

No

De Novo

All Spectra

DB peptides De novo only

PEAKS DB Workflow

De novo both helps to improve DB search, and reports novel peptides.

3. Three essential examinations to ensure result quality.

Don’t Trust Software Blindly!• Google “Don’t trust software blindly” returned

5,140,000 results.• As you quality control your experiments,

quality control the software’s results too.

Essential Examination 1

#decoy #targetin low score region

Low #decoy in high score region

Essential Examination 2

High scoring peptidesshould have low precursor error.

Precursor error start to scatterbelow threshold

Essential Examination 3• Spectrum annotation around score threshold.

Take Home Message

• Another year of dedicated work on PEAKS.• Ensured accuracy; maximized sensitivity.• Do the three essential examinations.– They are simple … at least in PEAKS.

“a big step forward” – Christian Schmelzer, Martin Luther University

Enjoy!

http://www.bioinfor.com/peaks-download-a-pricing