Table 1.

Summary of different SolidBin modes

SolidBin mode	Constraints	Type	Parameter	Description	Performance profiles
SolidBin-naive	None	Taxonomy-independent	None	The NCut mode	Comparable performance to other binners
SolidBin-SFS	ML	Taxonomy-independent	α	Take the contig pairs with high similarity as ML constraints	Good performance when the number of samples is large. It can obtain good results with high similarity quality, and more samples could bring more useful coverage information
SolidBin-coalign	ML	Taxonomy-dependent	α	Take the contig pairs with the same assignment by TAXAassign as ML constraints	Good performance when the genomes contained in the datasets are on the species level, while using the results from TAXAassign. It can obtain good results if the pairwise ML constraints have high accuracy
SolidBin-CL	CL	Taxonomy-dependent	β	Take the contig pairs assigned to different genera by TAXAassign as CL constraints	Good performance when the genomes contained in the datasets are on the species level, often worse than SolidBin-coalign, while using the results from TAXAassign. It can obtain good results if the pairwise CL constraints have high accuracy
SolidBin-SFS-CL	ML	Taxonomy-dependent	α	Remove the contig pairs assigned to different genera by TAXAassign from the constraints set on the basis of SolidBin-SFS	Often a little better than SolidBin-SFS at a cost of spending much time on taxonomy alignment