Smearing effect in Projected Orthogonalized Chemical Encounter Monitoring (POCHEMON)

Supervisor: Brigitte Geurts, MSc. and Mike Koeman, MSc.

POCHEMON is a new in-house developed method for multivariate data analysis specifically for micro-organismal interspecies interactions. The biochemistry associated with this kind of interaction is an extremely rich source for bioactive molecules (metabolites) that can be monitored by e.g. GC-MS. Analysis of the data is complicated by the fact that the species do not only mix, but also interact – which leads to novel specific chemistry. POCHEMON is a method extended from Principal Component Analysis (PCA) which allows us to separate the mixture of metabolites from the separated species from the metabolites associated to the interspecies interaction.

POCHEMON is a multi-step method:

  1. PCA on both monocultures à Mixing model
  2. Projection of the co-culture into this mixing model, and calculation of the co-culture residuals: those parts of the co-culture data not explained by mixing of the two mono-cultures.
  3. PCA on the residuals à competition model

B-internship BGeurts MKoeman fig 2015

This internship focuses on investigating the smearing effect in POCHEMON. When using the residuals of a PCA model for further analysis the residual space is orthogonal to the space of the PCA model. In practice this means that the mixing model already explains part of the variation of the interspecies interaction, and this can lead to the identification of the wrong variables in when the residuals are analyzed by a second PCA model. In this internship you will use two novel methods that should alleviate this effect and find out if these can improve the analysis of POCHEMON.