CellMix 1.6


Algorithms available in CellMix

Algorithms available in CellMix

The CellMix package includes several deconvolution algorithms, which differ in term of input and output data. The following table helps choosing an appropriate algorithm according to the data available and the desired output.

Description Basis Coef Marker Iter
lsfit Partial deconvolution of proportions using least-squares fits (Abbas et al. (2009))
- -
cs-lsfit Partial deconvolution of cell signatures using least-squares fits
- -
qprog Estimates proportions from known expression signatures using quadratic programming (Gong et al. (2011))
- -
cs-qprog Estimates constrained cell-specific signatures from proportions using quadratic programming [experimental]
- -
DSA Complete deconvolution using Digital Sorting Algorithm (Zhong et al. (2013))
-
csSAM Estimates cell/tissue specific signatures from known proportions using SAM (Shen-Orr et al. (2010))
-
DSection Estimates proportions from proportions priors using MCMC (Erkkila et al. (2010))
500
ssKL Semi-supervised NMF algorithm for KL divergence, using marker genes (Gaujoux et al. (2011))
3000
ssFrobenius Semi-supervised NMF algorithm for Euclidean distance, using marker genes (Gaujoux et al. (2011))
3000
meanProfile Compute proportion proxies as mean expression profiles -
-
deconf Alternate least-square NMF method, using heuristic constraints (Repsilber et al. (2010))
- 1000
 Required input 
 Estimated output 
 Required input and estimated output
BasisCell-specific signatures
CoefCell proportions
MarkerInput: cell-specific marker list
Output: cell-specific differential expression (e.g., Case vs. Control)

Other algorithms not – yet – available in CellMix

  • TEMT: A mixture model for expression deconvolution from RNA-seq in heterogeneous tissues (Li et al. (2013))
  • DeMix: Deconvolution for Mixed Cancer Transcriptomes Using Raw Measured Data (Ahn et al. (2013))
  • ISOpure: Computational purification of individual tumor gene expression profiles leads to significant improvements in prognostic prediction (Quon et al. (2013))
  • Statistical expression deconvolution from mixed tissue samples (Clarker et al. (2010))