Using Random Peptide Phage Display Libraries for

early Breast cancer detection

Ekaterina Nenastyeva

OUTLINE• Introduction

– Motivation for early cancer detection– State of the art– Proposed assay based on Random Peptide Phage Display Libraries and Next Generation

sequencing

• Data Set– Data preprocessing

• Approaches for early Breast cancer detection– Identification of peptides specific for Breast cancer– Discrimination based on the whole peptide library

• Results and evaluation– LOO cross-validation– Permutation test

• Future work– Enriching library by cancer specific peptides– PCA

Motivation for early cancer detection

• Earlier stages

Simpler/ more effective treatment• Promising earlier stage biomarkers: Antibodies

State of the artThe current methods of analysis of antitumor humoral immune response:– SEREX – SERPA– ELISA– Antigen microarrays– Random peptide microarrays

FRDK

cE

PADQV

NP

RYLAC

EF

W

Any antigen can be substituted by a library of random peptides

Phage envelop

Phage DNA

Peptide coding

sequence

PeptideA peptide sequence can mimic the epitope recognized by an antibody

Detailed assay

Data Set

6103 *

710*]4.6..4.1[

10 samples:– 5 cases = stage 0 breast cancer patients – 5 controls = cancer-free women

Each sample = 2 replicas

Each replica has– Number of distinct 7-mer peptides – Total number of peptides in a replica:

normalization Total number of distinct 7-mer peptides in all

replicas 710*5

810

controls cases

• Identification of peptides specific for Breast cancer

• Discrimination based on the whole list of peptides

Approaches for early Breast cancer detection

Discrimination based on specific peptides

MAX < MIN

controls cases

• Cancer specific peptides:

• Control specific peptides:

MIN > MAX

controls cases

Peptides specific for Breast cancer7-mers: 1; 6-mers: 9; 5-mers: 44 (There are no control specific peptides!)

Permutation test for discrimination based on specific peptides

Hypothesis: “Controls do not have any peptide distinguishing them from cases, and cases have no less than one 7-mer, nine 6-mer and forty four 5-mer specific peptides”

Permutation test:• permutations • P-value = 0.028

252510 С

AVG correlation:

Threshold : (0.12+0.03)/2=0.075

Discrimination based on the whole peptide library

Correlation between peptides assigned to cases is higher than

between controls0.03control case

0.12 case case

IF AVG correlation: caseOTHERWISE control

Thresholdunknown case

Leave-one-out cross-validation for discrimination based on correlation

• Sensitivity =0.8 (4/5 correct predicted cases)• Specificity =1 (5/5 correct predicted controls)• Accuracy = 0.9

Permutation test for leave-one-out • permutations• 5 permutations have accuracy 0.9

(including true statuses arrangement)• P-value = 0.02

252510 С

controlsA,B,C,E,H

casesD,F,G,I,J

Conclusion

•Discrimination method based on whole peptide library and correlation showed statistically significant results

•Found Breast cancer specific peptides were not statistically significant although the hypothesis that there were no peptides specific for controls was statistically significant

Future work

Discrimination methods based on:• Correlation and enriching library by cancer

specific peptides• Principal component analysis