Predicting Type III Effector Proteins Using the Effectidor Web Server.

Journal: Methods in molecular biology (Clifton, N.J.)
PMID:

Abstract

Various Gram-negative bacteria use secretion systems to secrete effector proteins that manipulate host biochemical pathways to their benefit. We and others have previously developed machine-learning algorithms to predict novel effectors. Specifically, given a set of known effectors and a set of known non-effectors, the machine-learning algorithm extracts features that distinguish these two protein groups. In the training phase, the machine learning learns how to best combine the features to separate the two groups. The trained machine learning is then applied to open reading frames (ORFs) with unknown functions, resulting in a score for each ORF, which is its likelihood to be an effector. We developed Effectidor, a web server for predicting type III effectors. In this book chapter, we provide a step-by-step introduction to the application of Effectidor, from selecting input data to analyzing the obtained predictions.

Authors

  • Naama Wagner
    The Shmunis School of Biomedicine and Cancer Research, George S. Wise Faculty of Life Sciences, Tel Aviv University, Tel Aviv, Israel.
  • Doron Teper
    Department of Molecular Biology and Ecology of Plants, Tel Aviv University, Tel Aviv, 69978, Israel.
  • Tal Pupko
    Department of Earth and Planetary Science, UC Berkeley, Berkeley, CA, 94720, USA.