Package Canopy. April 8, 2017

Transcription

1 Type Package Package Canopy April 8, 2017 Title Accessing Intra-Tumor Heterogeneity and Tracking Longitudinal and Spatial Clonal Evolutionary History by Next-Generation Sequencing Version Author Yuchao Jiang, Nancy R. Zhang Maintainer A statistical framework and computational procedure for identifying the sub-populations within a tumor, determining the mutation profiles of each subpopulation, and inferring the tumor's phylogenetic history. The input are variant allele frequencies (VAFs) of somatic single nucleotide alterations (SNAs) along with allele-specific coverage ratios between the tumor and matched normal sample for somatic copy number alterations (CNAs). These quantities can be directly taken from the output of existing software. Canopy provides a general mathematical framework for pooling data across samples and sites to infer the underlying parameters. For SNAs that fall within CNA regions, Canopy infers their temporal ordering and resolves their phase. When there are multiple evolutionary configurations consistent with the data, Canopy outputs all configurations along with their confidence assessment. License GPL-2 Depends R (>= 3.3), ape, fields, pheatmap, scatterplot3d Imports grdevices, graphics, stats, utils URL NeedsCompilation no Repository CRAN Date/Publication :30:16 UTC R topics documented: addsamp AML

2 2 addsamp canopy.bic canopy.cluster canopy.cluster.estep canopy.cluster.mstep canopy.output canopy.plot canopy.post canopy.sample canopy.sample.cluster canopy.sample.cluster.nocna canopy.sample.nocna getclonalcomposition getcmcm getcz getlikelihood getlikelihood.sna getq getvaf getz initialcna initialcnacopy initialp initialsna MDA MDA231_sampchain MDA231_ sampcna sampcnacopy sampp sampsna sampsna.cluster sortcna toy toy toy Index 32 addsamp To determine whether the sampled will be accepted To determine whether the sampled will be accepted by comparing the likelihood, used in canopy.sample.

3 AML43 3 addsamp(,.new).new input (current) input (newly sampled) returned (either retain the old or accept the new ). sna.name = MDA231$sna.name Y = MDA231$Y C = MDA231$C R = MDA231$R X = MDA231$X WM = MDA231$WM Wm = MDA231$Wm epsilonm = MDA231$epsilonM epsilonm = MDA231$epsilonm # sampling location of SNAs.new = MDA231_.new$sna = sampsna(mda231_).new$z = getz(.new, sna.name).new$q = getq(.new, Y, C).new$H =.new$q.new$vaf = getvaf(.new, Y).new$likelihood = getlikelihood(.new, R, X, WM, Wm, epsilonm, epsilonm) = addsamp(mda231_,.new) AML43 SNA input for primary tumor and relapse genome of leukemia patient from Ding et al. Nature SNAs from sequencing of leukemia patient at two timepoints. All SNAs are filtered to be from copy-number-neutral region.

4 4 canopy.bic data(aml43) List of simulated SNA input data for Canopy. data(aml43) canopy.bic To get BIC as a model selection criterion To get BIC as a model selection criterion from MCMC sampling results. canopy.bic(sampchain,projectname,k,numchain,burnin,thin,pdf) sampchain projectname K numchain burnin thin pdf list of sampled s returned by canopy.sample name of project number of subclones (vector) number of MCMC chains with random initiations burnin of MCMC chains MCMC chains thinning whether a pdf plot of BIC should be generated, default to be TRUE BIC values (vector) for model selection with plot generated (pdf format).

5 canopy.cluster 5 data(mda231_sampchain) sampchain = MDA231_sampchain projectname = 'MD231' K = 3:6 numchain = 20 burnin = 150 thin = 5 bic = canopy.bic(sampchain = sampchain, projectname = projectname, K = K, numchain = numchain, burnin = burnin, thin = thin) canopy.cluster EM algorithm for multivariate clustering of SNAs EM algorithm for multivariate clustering of SNAs. canopy.cluster(r, X, num_cluster, num_run, Mu.init = NULL, Tau_Kplus1 = NULL) R X num_cluster num_run Mu.init Tau_Kplus1 alternative allele read depth matrix total read depth matrix number of mutation clusters (BIC as model selection metric) number of EM runs for estimation for each specific number of clusters (to avoid EM being stuck in local optima) (optional) initial value of the VAF centroid for each mutation cluster in each sample (optional) pre-specified proportion of noise component in clustering, uniformly distributed between 0 and 1 Matrix of posterior probability of cluster assignment for each mutation.

6 6 canopy.cluster.estep data(aml43) R = AML43$R X = AML43$X Mu = AML43$Mu Tau = AML43$Tau pg = canopy.cluster.estep(tau, Mu, R, X) canopy.cluster.estep E-step of EM algorithm for multivariate clustering of SNAs E-step of EM algorithm for multivariate clustering of SNAs. Used in canopy.cluster. canopy.cluster.estep(tau, Mu, R, X) Tau Mu R X prior for proportions of mutation clusters MAF centroid for each mutation cluster in each sample alternative allele read depth matrix total read depth matrix Matrix of posterior probability of cluster assignment for each mutation. data(aml43) R = AML43$R X = AML43$X Mu = AML43$Mu Tau = AML43$Tau pg = canopy.cluster.estep(tau, Mu, R, X)

7 canopy.cluster.mstep 7 canopy.cluster.mstep M-step of EM algorithm for multivariate clustering of SNAs M-step of EM algorithm for multivariate clustering of SNAs. Used in canopy.cluster. canopy.cluster.mstep(pg, R, X, Tau_Kplus1) pg R X Tau_Kplus1 matrix of posterior probability of cluster assignment for each mutation alternative allele read depth matrix total read depth matrix proportion mutation cluster that is uniformly distributed to capture noise List of bic, converged Mu, Tau, and SNA cluster assignment. data(aml43) R = AML43$R; X = AML43$X num_cluster = 4 # Range of number of clusters to run num_run = 6 # How many EM runs per clustering step Tau_Kplus1=0.05 # Proportion of noise component Mu.init=cbind(c(0.01,0.15,0.25,0.45),c(0.2,0.2,0.01,0.2)) # initial value # of centroid canopy.cluster=canopy.cluster(r = R, X = X, num_cluster = num_cluster, num_run = num_run, Mu.init = Mu.init, Tau_Kplus1=Tau_Kplus1)

8 8 canopy.output canopy.output To generate a posterior To generate a posterior from the sub- space of s with the same configurations. canopy.output(post, config.i, C) post config.i C list returned by canopy.post configuration of sub- space to be output CNA and CNA-region overlapping matrix, only needed if overlapping CNAs are used as input posterior from the sub- space of s with the same configurations. data(mda231_sampchain) sampchain = MDA231_sampchain projectname = 'MD231' K = 3:6 numchain = 20 burnin = 150 thin = 5 optk = 4 C = MDA231$C post = canopy.post(sampchain = sampchain, projectname = projectname, K = K, numchain = numchain, burnin = burnin, thin = thin, optk = optk, C = C) config.i = 3 output. = canopy.output(post = post, config.i = config.i, C = C)

9 canopy.plot 9 canopy.plot To plot inferred by Canopy To plot Canopy s reconstructed phylogeny. Major plotting function of Canopy. canopy.plot(, pdf, pdf.name, txt, txt.name) pdf pdf.name txt txt.name input to be plotted whether a pdf plot should be generated, default to be FALSE name of pdf to be generated, has to be provided if pdf is to be generated whether a txt file should be generated with information on mutations along the branches, default to be FALSE name of txt to be generated, has to be provided if txt is to be generated Plot of structure, clonal frequency and mutation legends (pdf format). canopy.plot(mda231_, pdf = TRUE, pdf.name = 'MDA231_.pdf') canopy.post Posterior evaluation of MCMC sampled s Burnin, thinning, and posterior evaluation of MCMC sampled s. canopy.post(sampchain, projectname, K, numchain, burnin, thin, optk, C, post.config.cutoff)

10 10 canopy.post sampchain projectname K numchain burnin thin optk C list of sampled s returned by canopy.sample name of project number of subclones (vector) number of MCMC chains with random initiations burnin of MCMC chains MCMC chain thinning. optimal number of subclones determined by canopy.bic CNA and CNA-region overlapping matrix, only needed if overlapping CNAs are used as input post.config.cutoff cutoff value for posterior probabilities of configurations, default is set to be 0.05 (only configurations with greater than 0.05 posterior probabilities will be reported by Canopy) sampthin list of sampled posterior s sampthin.lik vector of likelihood of sampled posterior s config vector of configuration of sampled posterior s (integer values) config.summary summary of configurations of sampled posterior s data(mda231_sampchain) sampchain = MDA231_sampchain projectname = 'MD231' K = 3:6 numchain = 20 burnin = 150 thin = 5 optk = 4 C = MDA231$C post = canopy.post(sampchain = sampchain, projectname = projectname, K = K, numchain = numchain, burnin = burnin, thin = thin, optk = optk, C = C)

11 canopy.sample 11 canopy.sample MCMC sampling in space To sample the posterior s. Major function of Canopy. canopy.sample(r, X, WM, Wm, epsilonm, epsilonm, C=NULL, Y, K, numchain, max.simrun, min.simrun, writeskip, projectname, cell.line=null, plot.likelihood=null) R X WM Wm epsilonm epsilonm C Y K numchain max.simrun min.simrun writeskip projectname alternative allele read depth matrix total read depth matrix observed major copy number matrix observed minor copy number matrix observed standard deviation of major copy number (scalar input is transformed into matrix) observed standard deviation of minor copy number (scalar input is transformed into matrix) CNA and CNA-region overlapping matrix, only needed if overlapping CNAs are used as input SNA and CNA-region overlapping matrix number of subclones (vector) number of MCMC chains with random initiations maximum number of simutation iterations for each chain minimum number of simutation iterations for each chain interval to store sampled s name of project cell.line default to be FALSE, TRUE if input sample is cell line (no normal cell contamination) plot.likelihood default to be TRUE, posterior likelihood plot generated for check of convergence and selection of burnin and thinning in canopy.post List of sampleed s in sub space with different number of subclones; plot of posterior likelihoods in each sub space generated (pdf format).

12 12 canopy.sample.cluster R = MDA231$R; X = MDA231$X WM = MDA231$WM; Wm = MDA231$Wm epsilonm = MDA231$epsilonM; epsilonm = MDA231$epsilonm C = MDA231$C Y = MDA231$Y K = 3:6 numchain = 20 projectname = 'MDA231' # sampchain = canopy.sample(r = R, X = X, WM = WM, Wm = Wm, epsilonm = epsilonm, # epsilonm = epsilonm, C = C, Y = Y, K = K, numchain = numchain, # max.simrun = 50000, min.simrun = 10000, writeskip = 200, # projectname = projectname, cell.line = TRUE, plot.likelihood = TRUE) canopy.sample.cluster MCMC sampling in space with pre-clustering of SNAs To sample the posterior s with pre-clustering step of SNAs. Major function of Canopy. canopy.sample.cluster(r, X, sna_cluster, WM, Wm, epsilonm, epsilonm, C=NULL, Y, K, numchain, max.simrun, min.simrun, writeskip, projectname, cell.line=null, plot.likelihood=null) R X sna_cluster WM Wm epsilonm epsilonm C alternative allele read depth matrix total read depth matrix cluster assignment for each mutation from the EM Binomial clustering algorithm observed major copy number matrix observed minor copy number matrix observed standard deviation of major copy number (scalar input is transformed into matrix) observed standard deviation of minor copy number (scalar input is transformed into matrix) CNA and CNA-region overlapping matrix, only needed if overlapping CNAs are used as input

13 canopy.sample.cluster.nocna 13 Y K numchain max.simrun min.simrun writeskip projectname SNA and CNA-region overlapping matrix number of subclones (vector) number of MCMC chains with random initiations maximum number of simutation iterations for each chain minimum number of simutation iterations for each chain interval to store sampled s name of project cell.line default to be FALSE, TRUE if input sample is cell line (no normal cell contamination) plot.likelihood default to be TRUE, posterior likelihood plot generated for check of convergence and selection of burnin and thinning in canopy.post List of sampleed s in sub space with different number of subclones; plot of posterior likelihoods in each sub space generated (pdf format). R = MDA231$R; X = MDA231$X WM = MDA231$WM; Wm = MDA231$Wm epsilonm = MDA231$epsilonM; epsilonm = MDA231$epsilonm C = MDA231$C Y = MDA231$Y K = 3:6 numchain = 20 projectname = 'MDA231' # sampchain = canopy.sample.cluster(r = R, X = X, sna_cluster=c(1,2,3,4), # WM = WM, Wm = Wm, epsilonm = epsilonm, # epsilonm = epsilonm, C = C, Y = Y, K = K, numchain = numchain, # max.simrun = 50000, min.simrun = 10000, writeskip = 200, # projectname = projectname, cell.line = TRUE, plot.likelihood = TRUE) canopy.sample.cluster.nocna MCMC sampling in space with pre-clustering of SNAs To sample the posterior s with pre-clustering step of SNAs. Major function of Canopy.

14 14 canopy.sample.cluster.nocna R X canopy.sample.cluster.nocna(r, X, sna_cluster, K, numchain, max.simrun, min.simrun, writeskip, projectname, cell.line=null, plot.likelihood=null) sna_cluster K numchain max.simrun min.simrun writeskip projectname alternative allele read depth matrix total read depth matrix cluster assignment for each mutation from the EM Binomial clustering algorithm number of subclones (vector) number of MCMC chains with random initiations maximum number of simutation iterations for each chain minimum number of simutation iterations for each chain interval to store sampled s name of project cell.line default to be FALSE, TRUE if input sample is cell line (no normal cell contamination) plot.likelihood default to be TRUE, posterior likelihood plot generated for check of convergence and selection of burnin and thinning in canopy.post List of sampleed s in sub space with different number of subclones; plot of posterior likelihoods in each sub space generated (pdf format). data(toy3) R = toy3$r; X = toy3$x sna_cluster = toy3$sna_cluster K = 3:5 numchain = 10 projectname = 'toy3' # sampchain = canopy.sample.cluster.nocna(r = R, X = X, # sna_cluster=sna_cluster, K = K, numchain = numchain, # max.simrun = 40000, min.simrun = 10000, writeskip = 200, # projectname = projectname, # cell.line = TRUE, plot.likelihood = TRUE)

15 canopy.sample.nocna 15 canopy.sample.nocna MCMC sampling in space To sample the posterior s without CNA input. Major function of Canopy. R X K numchain canopy.sample.nocna(r, X, K, numchain, max.simrun, min.simrun, writeskip, projectname, cell.line=null, plot.likelihood=null) max.simrun min.simrun writeskip projectname alternative allele read depth matrix total read depth matrix number of subclones (vector) number of MCMC chains with random initiations maximum number of simutation iterations for each chain minimum number of simutation iterations for each chain interval to store sampled s name of project cell.line default to be FALSE, TRUE if input sample is cell line (no normal cell contamination) plot.likelihood default to be TRUE, posterior likelihood plot generated for check of convergence and selection of burnin and thinning in canopy.post List of sampleed s in sub space with different number of subclones; plot of posterior likelihoods in each sub space generated (pdf format). data(toy3) R = toy3$r; X = toy3$x K = 3:5 numchain = 10 projectname = 'toy3' # sampchain = canopy.sample.nocna(r = R, X = X, K = K, numchain = numchain,

16 16 getcmcm # max.simrun = 50000, min.simrun = 10000, writeskip = 200, # projectname = projectname, # cell.line = TRUE, plot.likelihood = TRUE) getclonalcomposition To get clonal composition To get clonal composition (mutational profile of each clone) of. Used in canopy.post. getclonalcomposition() input List of each clone s mutational profile. getclonalcomposition(mda231_) getcmcm To get major and minor copy per clone To get major and minor copy per clone. Used in canopy.sample. getcmcm(, C) C input CNA regions and CNA overlapping matrix

17 getcz 17 CM Cm Matrix of major copy per clone. Matrix of minor copy per clone. C = MDA231$C getcmcm(mda231_, C) getcz To get CNA genotyping matrix CZ To get CNA genotyping matrix CZ from location of CNAs on the. Used in canopy.sample. getcz() input CNA genotyping matrix CZ. getcz(mda231_)

18 18 getlikelihood getlikelihood To get likelihood of the To get likelihood of the given struture and data input. Used in canopy.sample. getlikelihood(,r,x,wm,wm,epsilonm,epsilonm) R X WM Wm epsilonm epsilonm input alternative allele read depth matrix total read depth matrix observed major copy number matrix observed minor copy number matrix observed standard deviation of major copy number (scalar input is transformed into matrix) observed standard deviation of minor copy number (scalar input is transformed into matrix) Likelihood of sampled. R = MDA231$R X = MDA231$X WM = MDA231$WM Wm = MDA231$Wm epsilonm = MDA231$epsilonM epsilonm = MDA231$epsilonm getlikelihood(mda231_, R, X, WM, Wm, epsilonm, epsilonm)

19 getlikelihood.sna 19 getlikelihood.sna To get SNA likelihood of the To get SNA likelihood of the given struture and data input. Used in canopy.sample.nocna and canopy.sample.cluster.nocna. getlikelihood.sna(, R, X) R X input alternative allele read depth matrix total read depth matrix Likelihood of sampled. R = MDA231$R X = MDA231$X getlikelihood.sna(mda231_, R, X) getq To get SNA-CNA genotyping matrix To get SNA-CNA genotyping matrix Q, which specifies whether an SNA precedes a CNA. Used in canopy.sample. getq(, Y, C)

20 20 getvaf Y C input SNA CNA overlapping matrix CNA and CNA region overlapping matrix Genotyping matrix Q. Y = MDA231$Y C = MDA231$C getq(mda231_, Y, C) getvaf To get variant allele frequency (VAF) To get variant allele frequency (VAF) matrix, which contains percentage of mutant SNA alleles across samples. Used in canopy.sample. getvaf(,y) Y input SNA CNA overlapping matrix Variant allele frequency matrix VAF.

21 getz 21 Y = MDA231$Y getvaf(mda231_, Y) getz To get SNA genotyping matrix Z To get SNA genotyping matrix Z from location of SNAs on the. Used in canopy.sample. getz(, sna.name) sna.name input vector of SNA names Genotyping matrix Z. sna.name = rownames(mda231$r) getz(mda231_, sna.name)

22 22 initialcnacopy initialcna To initialize positions of CNAs To initialize positions of CNAs on the. Used in initialization step of canopy.sample. initialcna(,cna.name) cna.name input vector of input CNA names Matrix specifying positions of CNAs (start and end node). cna.name = rownames(mda231$wm) initialcna(mda231_, cna.name) initialcnacopy To initialize major and minor copies of CNAs To initialize major and minor copies of CNAs. Used in initialization step of canopy.sample. initialcnacopy() input

23 initialp 23 Matrix specifying major and minor copies of CNAs. initialcnacopy(mda231_) initialp To initialize clonal frequency matrix To initialize clonal frequency matris P. Used in initialization step of canopy.sample. initialp(,sampname,cell.line) sampname cell.line input vector of input sample names default to be FALSE, TRUE if input sample is cell line (no normal cell contamination) Clonal frequency matrix P. sampname = colnames(mda231$r) initialp(mda231_, sampname, cell.line = TRUE)

24 24 MDA231 initialsna To initialize positions of SNAs To initialize positions of SNAs on the. Used in initialization step of canopy.sample. initialsna(,sna.name) sna.name input vector of input SNA names Matrix specifying positions of SNAs (start and end node). sna.name = rownames(mda231$r) initialsna(mda231_, sna.name) MDA231 Dataset for project MDA231 Pre-stored dataset for project MDA231. A transplantable metastasis model system was derived from a heterogeneous human breast cancer cell line MDA-MB-231. Cancer cells from the parental line MDA-MB-231 were engrafted into mouse hosts leading to organ-specific metastasis. Mixed cell populations (MCPs) were in vivo selected from either bone or lung metastasis and grew into phenotypically stable and metastatically competent cancer cell lines. The parental line as well as the MCP sublines were whole-exome sequenced with somatic SNAs and CNAs profiled.

25 MDA231_sampchain 25 List of input data for Canopy from project MDA231. MDA231_sampchain List of pre-sampled s List of sampleed s in sub space with different number of subclones for project MDA231. data(mda231_sampchain) List of sampled s from different sub space data(mda231_sampchain)

26 26 sampcna MDA231_ Most likely from project MDA231 Most likely from project MDA231 as a example. Most likely from project MDA231 sampcna To sample CNA positions To sample CNA positions along the. Used in canopy.sample. sampcna() input Newly sampled matrix specifying positions of CNAs (start and end node).

27 sampcnacopy 27 sampcna(mda231_) sampcnacopy To sample major and minor copies of CNAs To sample major and minor copies of CNAs. Used in canopy.sample. sampcnacopy() input Newly sampled matrix specifying major and minor copies of CNAs. sampcnacopy(mda231_) sampp To sample clonal frequency To sample clonal frequency matrix P. Used in canopy.sample. sampp(, cell.line)

28 28 sampsna cell.line input default to be FALSE, TRUE if input sample is cell line (no normal cell contamination) Newly sampled clonal frequency matrix P. sampp(mda231_, cell.line = TRUE) sampsna To sample SNA positions To sample SNA positions along the. Used in canopy.sample. sampsna() input Newly sampled matrix specifying positions of SNAs (start and end node). sampsna(mda231_)

29 sampsna.cluster 29 sampsna.cluster To sample positions of SNA clusters To sample SNA cluster positions along the. Used in canopy.sample.cluster and canopy.sample.cluster.nocna. sampsna.cluster() input Newly sampled matrix specifying positions of SNA clusters (start and end node). MDA231_$sna.cluster=initialsna(MDA231_,paste('cluster',1:4,sep='')) sampsna.cluster(mda231_) sortcna To sort identified overlapping CNAs. To sort identified overlapping CNAs by their major and minor copy numbers. Used in canopy.post. sortcna(,c) C input CNA and CNA-region overlapping matrix

30 30 toy Tree whose overlapping CNAs are sorted by major and minor copy numbers. C = MDA231$C sortcna(mda231_, C) toy Toy dataset for Canopy Pre-stored simulated toy dataset. data(toy) List of simulated input data for Canopy. data(toy)

31 toy2 31 toy2 Toy dataset 2 for Canopy Pre-stored simulated toy dataset. data(toy2) List of simulated input data for Canopy. data(toy2) toy3 Toy dataset 3 for Canopy Pre-stored simulated toy dataset. 200 simulated SNAs from a with 4 branches. No CNA events at play. data(toy3) List of simulated SNA input data for Canopy. data(toy3)

32 Index Topic datasets AML43, 3 MDA231, 24 MDA231_sampchain, 25 MDA231_, 26 toy, 30 toy2, 31 toy3, 31 Topic package addsamp, 2 canopy.bic, 4 canopy.cluster, 5 canopy.cluster.estep, 6 canopy.cluster.mstep, 7 canopy.output, 8 canopy.plot, 9 canopy.post, 9 canopy.sample, 11 canopy.sample.cluster, 12 canopy.sample.cluster.nocna, 13 canopy.sample.nocna, 15 getclonalcomposition, 16 getcmcm, 16 getcz, 17 getlikelihood, 18 getlikelihood.sna, 19 getq, 19 getvaf, 20 getz, 21 initialcna, 22 initialcnacopy, 22 initialp, 23 initialsna, 24 sampcna, 26 sampcnacopy, 27 sampp, 27 sampsna, 28 sampsna.cluster, 29 sortcna, 29 addsamp, 2 AML43, 3 canopy.bic, 4, 10 canopy.cluster, 5, 6, 7 canopy.cluster.estep, 6 canopy.cluster.mstep, 7 canopy.output, 8 canopy.plot, 9 canopy.post, 8, 9, 11, 13 16, 29 canopy.sample, 2, 4, 10, 11, 16 24, canopy.sample.cluster, 12, 29 canopy.sample.cluster.nocna, 13, 19, 29 canopy.sample.nocna, 15, 19 getclonalcomposition, 16 getcmcm, 16 getcz, 17 getlikelihood, 18 getlikelihood.sna, 19 getq, 19 getvaf, 20 getz, 21 initialcna, 22 initialcnacopy, 22 initialp, 23 initialsna, 24 MDA231, 24 MDA231_sampchain, 25 MDA231_, 26 sampcna, 26 sampcnacopy, 27 sampp, 27 sampsna, 28 sampsna.cluster, 29 sortcna, 29 toy, 30 32

33 INDEX 33 toy2, 31 toy3, 31