Bioinformatic analysis of peptide microarray immunoassay data for serological diagnosis of infectious diseases

  • Kate Nambiar

Student thesis: Doctoral Thesis


Understanding antibody - antigen interactions occurring in infectious diseases
is important in understanding aetiology, can help facilitate diagnosis, and could
offer potential targets for vaccine or therapeutic antibody development. Peptide
arrays – collections of short peptides immobilised on solid planar supports –
offer a high throughput and highly parallel method of identifying immunogenic
epitopes and relating patterns of antibody identification to clinical disease states.
As technology advances, so the density and complexity of peptide arrays of
becomes ever higher. Managing the large volume of data that modern high
density microarrays generate requires sophisticated bioinformatics in order to
minimise errors and biases.

In this thesis I introduce a new software package, pmpa, that uses R, the open
source statistical programming platform and an object orientated framework
from the Bioconductor project. The package facilitates analysis of peptide microarray
data including functions for reading scanned data files, quality assessment
and pre-processing. It is both flexible and modular – integrating with
existing software in the Bioconductor repository.

Data pre-processing is key to any microarray analysis. Noise due to technical
variation can obscure true biological effects if careful steps are not taken. The aim
of pre-processing is to minimise noise while preserving biological variation. No
consensus exists as to the optimal method of pre-processing making comparison
between studies difficult. This thesis explores two key aspects of pre-processing:
background correction and normalisation using two experimental datasets – a
titration series of a monoclonal anti C.difficle Toxin B monoclonal antibody, and
dataset with an anti-Toxin A antibody spiked into non immune sera to examine
biases introduced by the pre-processing and whether they improve measures such
as precision and differential identification.

Finally the analysis method is applied to two studies identifying antibody signatures
in infectious diseases: the first investigating immune responses to C. difficile
– a major hospital acquired infection and the leading identifiable cause of antibiotic
associated diarrhoea, and the second characterising antibody signatures
that define paediatric tuberculosis infection. The real world application of the
methodology identifies signatures of immune responses characterising clinical
disease eg. relapsing vs. single episode C. difficile infection, but also highlights a
number of limitations of the technique such as batch confounding and response
Date of AwardApr 2017
Original languageEnglish
Awarding Institution
  • University of Brighton
SupervisorMartin Llewelyn (Supervisor)

Cite this