- Changes to prepare package for submission to Bioconductor. (No functional changes)
- Fix bug in devianceIC() = D_post + 2 * p_D; and add both the original p_D, i.g., D_mean - D_post and the Gelman's alternative p_D = 2 * var(D). By default, DIC uses Gelman's method.
- Suggest narrowing theta1 for cardelino-free, e.g., prior1=c(45, 55)
- Add get_logLik() function to get log likelihood for clone_id_Gibbs(), namely P(A, D | C, I, theta0, theta1)
- Add devianceIC() for Deviance information criterion (DIC) for clone_id_Gibbs()
- Change the default of relax_Config=TRUE for clone_id() and clone_id_Gibbs()
- Fix the bug in
Geweke_Z
function for convergence diagnostic - Change the default number of iterations in
clone_id
to 5000 - Add more tests
- Change the default colMatch to greedy search
- Rename the cell_assign_Gibbs to clone_id_Gibbs
- Rename the cell_assign_EM to clone_id_EM
- Move get_tree from clone_id.R to tree_utils.R
- Remove "Bernoulli" model in both clone_id_EM and clone_id_Gibbs. The user can transfer the binomial to Bernoulli by setting a threshold beforehand
- Remove the A_germ and D_germ parameters, leave it to future development to jointly modelling the germline variants
- Remove function load_vcf_h5 and dependency hdf5r
- Remove donor id relevant information and backup donor_id into
v0.4.2
and a new branchwith_vireo
- Remove donor_id.R, bin/run_vireo.R
- Remove tests/testthat/test-donorid.R, vignette/vignette-vireo.Rmd
- Remove unnecessary logo photos in inst
- Add binaryROC function of ROC curve
- Add message for switch vireo to Python implementation
- Support direct loading sparse matrix from cellSNP file in hdf5 format
- Support multiple chains when running clone_id (one chain by default)
- Support relabel during Gibbs sampling when running clone_id (not in use by default)
- Change sim_read_count: the default of wise0 is to "element", i.e., sequencing error rate is both variant and cell specific. Also, a new parameter sample_cell is added for the option to turn off sampling cells from seed D
- Change heat_matrix for supporting rownames and colnames matching the original order in the matrix instead of alphabetical order
- Add a runnable R file: run_vireo.R in /bin to run from command line
- Change load_GT_vcf by supporting GP for genotype probability
- Change load_cellSNP_vcf by supporting GL for genotype likelihood
- Change heat_matrix to for scale_x_discrete by default
- Change the convergence diagnosis in cell_assign_Gibbs to all cells
- Change the donor_read_simulator to beta-binomial for each variant in each cell
- Fix a bug in colMatch in
force
mode - Specify the fixed relax rate in cell_assign_Gibbs, which can be used to turn the input Config as a uniform prior when setting relax_rate_fixed=0.5
- Add averaged
Config_prob
andrelax_rate
in cell_assign_Gibbs outputs - Add
force
option in colMatch function to force one-to-one match
- Fix the bug in inferring the relax_rate in cell_assign_Gibbs
- Supporting learning the relax rate on clone configuration automatically
- fix the clone id missing for prob_mat.
- change the default the sampling iteration to 3000.
- cell_assign_Gibbs supports updating clone Configuration now. Set relax_Config between 0 and 1.
- change the default parameters for beta prior to better represent allelic dropout and imbalance in scRNA-seq data. It involves functions: sim_read_count, donor_read_simulator, and cell_assign_Gibbs
- minor change of pub.theme: title will be plain rather than bold.
- vireo supports match SNP from donor_data to cell_data; change default number of processors to n_proc=1
- change load_cellSNP_vcf default parameters to support more general case
- add more dependency to pass tests
- correct test-donor_id.R
- remove vignette-donorid.Rmd vignette and correct vignette-vireo.Rmd
- change donor_id to vireo
- fix minor bug for n_vars in vireo (i.e., donor_id)
- add vignette for demultiplexing without genotype
- change assessment for doublet detection indicator to prob_doublet
- add examples and remove unnecessary functions to pass biocCheck