Consensus Rules for Amino-Acid Characterisation using a Kernel method

Principle of use

  CRAACK submits a formated imput file  to the five following typing softwares:

Typing Tool
Rescue 1 Neural Network
Pons et al (1999), J Biomol NMR.
Platon Discriminant Analysis
Labudde et al(2003). J Biomol NMR.
Rescue 1N Neural Network Auguin et al (2003), Spectroscopy.
Rescue 2 Bayesian Algorithm
Marin et al. (2003).J.Biomol.NMR.
SVMTyp Support Vector Machine *
JL Pons - not published

* SVMTyp was built from SVMtorch (Ronan Collobert et al (2001) Journal of Machine Learning Research)

CRAACK is based on two consensus algorithm which tries to acccurate the responses of existant typing modules.
The first algorithm accomplishes the analysis with SVM technologie ( MySVM, Joachims, Thorsten (1999) Support Vector Learning, chapter 11. MIT Press).
The second one use a classical algorithm of vote.
Additionaly, secondary structural prediction is available (CSI :Wishart & Sykes. (1994) . J Biomol NMR)

Fig: Craack general flowchart:

CRAACK proposes a page containing the results of the softwares and a consensus  result tab :

Num	Modules 	 Results	

1 [Input Shifts]: (8.264; 115.78; 4.410; 4.263; 62.34; 70.01; 174.11; #1; T; )
[Rescue1]: G (67)
[Rescue1N]: T (39)
[Rescue2]: T (1.00)
[Platon]: T (56.22) S (44.55)
[SVMTyping]: T (5.33)

[SVM Consensus]: T (7.34)
[Vote Consensus]:T (5.4) G (1) S (0.4)
[CSI]: E(-1) E(-1) E(-1)

Download Craack (Linux version)

Authors & Reference

Benod C., Delsuc MA., Pons JL (2006) J. Chem. Inf. Model. 46 (3) 1517-1522