Table 1:

A summary of the available applications used for base-calling on the Illumina platform

Name	Statistical approach	Biases explicitly corrected	Training data required	Quality score	Practical notes	References
Bustard	Parametric Model	Σ, ϕ, δ	No	Phred	Not freely retrievable
Alta-Cyclic	Mixed Parametric and SVM	Σ, ϕ, δ	Yes	Phred	No longer maintained; requires a Sun Grid Engine cluster environment	[16]
Rolexa	Parametric Model	Σ, ϕ, ω	No	IUPAC	No longer maintained	[17]
Swift	Parametric Model	Σ, ϕ, µ	No	Phred	No longer maintained	[15]
BayesCall/ naiveBayesCall	Parametric Model	Σ, ϕ, δ	No	Phred		[19, 22]
Seraphim	Parametric Model	Σ, ϕ, δ	No	Phred	We did not succeed installing it	[21]
Ibis	Fully empirical SVM	(n/a)	Yes	Phred		[18]
BING	Parametric Model	Σ, ϕ	No	None	Not freely retrievable; requires own image processing as input	[23]

We give a short description of the statistical approach used by each application. Next, the biases explicitly modeled and corrected by the application are reported (see Figure 1 for details). Alta-cyclic and Ibis rely on supervised learning and require training data. Finally uncertainty measurements or sequencing quality is either reported as Phred scores or using IUPAC codes. For details, please refer to the main text.