@@ -47,7 +47,9 @@ Suggests:

     ggdendro,

     broom,

     lme4,

-    RColorBrewer

+    RColorBrewer,

+    SingleCellExperiment,

+    scater

 LinkingTo: 

     Rcpp

 VignetteBuilder: 

@@ -1,5 +1,6 @@

-# Todo

-# combine methods

+# Contig -- always equalize to cell

+# Cluster -- equalize to contig

+# Cell -- equalize after calling

 valid_KeyedTbl = function(tbl, keys){

     tbl_nm = deparse(substitute(tbl))

@@ -25,6 +26,7 @@ valid_KeyedTbl = function(tbl, keys){

 #' @param cell_pk character vector naming fields in `cell_tbl` that uniquely identify a cell barcode

 #' @param cluster_tbl A data frame that provide cluster assignments for each contig

 #' @param cluster_pk If `cluster_tbl` was provided, a character vector naming fields in `cluster_tbl` that uniquely identify a cluster

+#' @param equalize `logical`. Should the contig, cells and clusters be equalized by taking the intersection of their common keys?

 #' @return \code{ContigCellDB}

 #'

 #' @section Accessors/mutators:

@@ -82,9 +84,10 @@ ContigCellDB = function(contig_tbl, contig_pk, cell_tbl, cell_pk, cluster_tbl, c

 }

 #' @describeIn ContigCellDB-fun provide defaults that correspond to identifiers in 10X VDJ data

+#' @param ... passed to [ContigCellDB()]

 #' @export

-ContigCellDB_10XVDJ = function(contig_tbl, contig_pk = c('barcode', 'contig_id'), cell_pk = 'barcode'){

-    ContigCellDB(contig_tbl = contig_tbl, contig_pk = contig_pk, cell_pk = cell_pk)

+ContigCellDB_10XVDJ = function(contig_tbl, contig_pk = c('barcode', 'contig_id'), cell_pk = 'barcode', ...){

+    ContigCellDB(contig_tbl = contig_tbl, contig_pk = contig_pk, cell_pk = cell_pk, ...)

 }

 access_cdb = function(x, name){

@@ -104,7 +107,7 @@ replace_cdb = function(x, name, value){

     }

     if(name %in% c('contig_tbl','contig_pk')){

         valid_KeyedTbl(x$contig_tbl, x$contig_pk)

-        x = equalize_ccdb(x)

+        x = equalize_ccdb(x, contig = TRUE, cell = FALSE, cluster = TRUE)

     }

     if(name %in% c('cell_tbl','cell_pk')){

         valid_KeyedTbl(x$cell_tbl, x$cell_pk)

@@ -112,7 +115,7 @@ replace_cdb = function(x, name, value){

     }

     if(name %in% c('cluster_tbl','cluster_pk')){

         valid_KeyedTbl(x$cluster_tbl, x$cluster_pk)

-        x = equalize_ccdb(x, contig = TRUE, cell = FALSE, cluster = FALSE)

+        x = equalize_ccdb(x, contig = TRUE, cell = FALSE, cluster = TRUE)

     }

     invisible(x)

 }

@@ -141,7 +144,7 @@ setMethod("$", signature = c(x = 'ContigCellDB'), access_cdb)

 #' ccdb_ex

 #' ccdb_ex$contig_tbl = dplyr::filter(ccdb_ex$contig_tbl, pop == 'b6')

 #' ccdb_ex

-setReplaceMethod("$", signature = c(x = 'ContigCellDB'), replace_cdb)

+setReplaceMethod("$", signature = c(x = 'ContigCellDB'), function(x, name, value) replace_cdb(x, name, value))

 setMethod('show', signature = c(object = 'ContigCellDB'), function(object){

     cat(class(object), "of", nrow(object$contig_tbl), "contigs")

@@ -200,20 +203,20 @@ equalize_ccdb = function(x, cell = TRUE, contig = TRUE, cluster = TRUE){

     if(cell) x@cell_tbl = semi_join(x$cell_tbl, x$contig_tbl, by = x$cell_pk)

     if(contig) x@contig_tbl = semi_join(x$contig_tbl, x$cell_tbl, by = x$cell_pk)

     if(nrow(x$cluster_tbl) > 0 && cluster) x@cluster_tbl = semi_join(x$cluster_tbl, x$contig_tbl, by = x$cluster_pk)

-    x@equalized = TRUE

+    x@equalized = (cell & contig & cluster) | x@equalized

     x

 }

 replace_cluster_tbl = function(ccdb, cluster_tbl, contig_tbl, cluster_pk){

     if(nrow(ccdb$cluster_tbl)>0 && !missing(cluster_pk)){

-        warning("Replacing `cluster_tbl` with ", paste(ccdb$cluster_pk, sep = ', ', '.'))

+        warning("Replacing `cluster_tbl` with ", paste(ccdb$cluster_pk, sep = ', '), '.')

     }

     if(!missing(cluster_pk)) ccdb@cluster_pk = cluster_pk

     ccdb@cluster_tbl = cluster_tbl

     if(!missing(contig_tbl)){

+        if(nrow(contig_tbl) != nrow(ccdb@contig_tbl)) stop("Length mismatch; this is a bug!")

         ccdb@contig_tbl = contig_tbl

         valid_KeyedTbl(ccdb$contig_tbl, ccdb$contig_pk)

-        ccdb = equalize_ccdb(ccdb, cell = FALSE, cluster = FALSE, contig = TRUE)

     }

     valid_KeyedTbl(ccdb$cluster_tbl, ccdb$cluster_pk)

     validObject(ccdb)

@@ -271,7 +274,7 @@ mutate_cdb <- function(ccdb, ..., tbl='contig_tbl'){

 #' splat = split_cdb(ccdb_ex, 'chain', 'contig_tbl')

 #' stopifnot(all(splat$TRA$contig_tbl$chain == 'TRA'))

 #' stopifnot(all(splat$TRB$contig_tbl$chain == 'TRB'))

-split_cdb = function(ccdb, fields, tbl = 'contig_tbl', drop = FALSE){

+split_cdb = function(ccdb, fields, tbl = 'contig_tbl', drop = FALSE, equalize = TRUE){

     thetbl = access_cdb(ccdb, tbl)

     if(!is.character(fields)){

         stop('`field` must be a character naming fields in `tbl`')

@@ -281,7 +284,8 @@ split_cdb = function(ccdb, fields, tbl = 'contig_tbl', drop = FALSE){

     }

     split_tbl = split(thetbl, thetbl[fields], drop = drop)

     out = purrr::map(split_tbl, function(tt){

-        replace_cdb(ccdb, tbl, tt)

+        tmp = replace_cdb(ccdb, tbl, tt)

+        if(equalize) equalize_ccdb(tmp)

     })

     out

 }

@@ -293,7 +297,7 @@ rbind.ContigCellDB <- function(..., deparse.level=1)

     .bind_rows_ccdb(objects[[1L]], objects[-1L])

 }

-#' #' Combine `ContigCellDB` along rows (contigs, cells or clusters)

+#' Combine `ContigCellDB` along rows (contigs, cells or clusters)

 #'

 #' The union of the rows in each of the objects is taken,

 #'  thus removing any rows that has an exact duplicate.  This

@@ -324,7 +328,5 @@ setMethod('rbind', 'ContigCellDB', rbind.ContigCellDB)

     for(tt in c('contig_pk', 'cell_pk', 'cluster_pk')){

         pks[[tt]] = purrr::reduce(purrr::map(all_objs, access_cdb, name = tt), union)

     }

-

-

     ContigCellDB(tbls$contig_tbl, pks$contig_pk, tbls$cell_tbl, pks$cell_pk, tbls$cluster_tbl, pks$cluster_pk)

 }

@@ -31,7 +31,7 @@ test_that('Can rbind', {

    expect_equal(unite, ccdb_ex)

    ccdb_ex = cluster_germline(ccdb_ex, segment_keys = 'sample')

-   splat_cell = split_cdb(ccdb_ex, c('sample', 'pop'), 'cell_tbl', drop = TRUE)

+   splat_cell = split_cdb(ccdb_ex, c('sample', 'pop'), 'cell_tbl', drop = TRUE, equalize = TRUE)

    unite_cell = do.call(rbind, splat_cell)

    expect_equal(unite_cell, ccdb_ex)

@@ -3,7 +3,7 @@ context('Cluster Logistic Testing')

 data(ccdb_ex)

 ccdb_ex = cluster_germline(ccdb_ex) %>% fine_clustering('cdr3_nt', 'DNA')

 ccdb_ex = canonicalize_cluster(ccdb_ex)

-ccdb_ex_trb = filter_cdb(ccdb_ex, chain == 'TRB') %>% canonicalize_cell(, contig_filter_args = chain == 'TRB', contig_fields = c('chain', 'v_gene', 'j_gene'))

+ccdb_ex_trb = filter_cdb(ccdb_ex, chain == 'TRB') %>% equalize_ccdb() %>% canonicalize_cell(, contig_filter_args = chain == 'TRB', contig_fields = c('chain', 'v_gene', 'j_gene'))

 theclust = filter(ccdb_ex$cluster_tbl, chain == "TRB", v_gene == "TRBV13-2", j_gene == "TRBJ2-7") %>% mutate( x_ = 1)

@@ -15,16 +15,18 @@ theform = ~ pop + (1|sample)

 test_that('Manual glmer testing equals cluster_logistic_test', {

     if(!requireNamespace('lme4')) skip('Install lme4')

     library(lme4)

-    # test against a manually data -- TRUE if the cell was in the cluster, false otherwise

+    # test against a manual data -- TRUE if the cell was in the cluster, false otherwise

     manual_glmer = glmer(update(theform, x_ ~ .) , data = manual_cell,  family = 'binomial')

     automatic1 = cluster_logistic_test(theform, ccdb = ccdb_ex_trb, cluster_whitelist = theclust, keep_fit = TRUE, silent = TRUE)

     expect_equal(fixef(automatic1$fit[[1]]), fixef(manual_glmer))

+    # Adding additional clusters doesn't change results

     wl2 = tibble(cluster_idx = c(theclust$cluster_idx, ccdb_ex_trb$cluster_tbl$cluster_idx[1:3]))

     automatic2 = cluster_logistic_test(theform, ccdb = ccdb_ex_trb, cluster_whitelist = wl2, keep_fit = TRUE, silent = TRUE)

     expect_equal(unique(automatic2$cluster_idx), wl2$cluster_idx)

     expect_equal(semi_join(automatic2, theclust, by = 'cluster_idx') %>% select(term:p.value), automatic1 %>% select(term:p.value))

+

     automatic3 = cluster_test_by(ccdb_ex, formula = theform, cluster_whitelist = wl2, silent = TRUE)

     expect_false(any(automatic3$chain == 'TRA'))

     expect_equal(semi_join(automatic3, theclust, by = 'cluster_idx') %>% select(term:p.value), automatic1 %>% select(term:p.value))

@@ -1,8 +1,8 @@

 ---

-title: "Combining Repertoire with Expression Using Nested colData"

+title: "Combining Repertoire with Expression with SingleCellExperiment"

 output: BiocStyle::html_document

 vignette: >

-  %\VignetteIndexEntry{Clustering repertoire via CDR3 sequences}

+  %\VignetteIndexEntry{Combining Repertoire with Expression with SingleCellExperiment}

   %\VignetteEngine{knitr::rmarkdown}

   %\VignetteEncoding{UTF-8}

 ---

@@ -15,7 +15,6 @@ knitr::opts_chunk$set(

 ```{r setup}

 library(CellaRepertorium)

-library(MultiAssayExperiment)

 library(SingleCellExperiment)

 library(dplyr)

 library(ggplot2)

@@ -62,7 +61,7 @@ sce = SingleCellExperiment(assay = expression, colData = all_bc)

 ```{r}

 ccdb2 = ContigCellDB(ccdb_ex$contig_tbl, contig_pk = ccdb_ex$contig_pk, cell_tbl = colData(sce), cell_pk = ccdb_ex$cell_pk, equalize = FALSE)

-ccdb2 = cdhit_ccdb(ccdb2, 'cdr3', type = 'AA', cluster_name = 'aa80', identity = .8)

+ccdb2 = cdhit_ccdb(ccdb2, 'cdr3', type = 'AA', cluster_name = 'aa80', identity = .8, min_length = 5)

 ccdb2 = fine_clustering(ccdb2, sequence_key = 'cdr3', type = 'AA', keep_clustering_details = FALSE)

 ```

@@ -73,9 +72,9 @@ Key is to construct with `equalize = FALSE`, which will allow some cells to lack

 # Chain pairings

 ```{r}

-colData(sce)$alpha =  canonicalize_cell(ccdb2, chain == 'TRA', contig_fields = c('chain', 'v_gene','d_gene', 'j_gene', 'AA97'))

+colData(sce)$alpha =  canonicalize_cell(ccdb2, chain == 'TRA', contig_fields = c('chain', 'v_gene','d_gene', 'j_gene', 'aa80'))

-colData(sce)$beta =  canonicalize_cell(ccdb2, chain == 'TRB', contig_fields = c('chain', 'v_gene','d_gene', 'j_gene', 'AA97'))

+colData(sce)$beta =  canonicalize_cell(ccdb2, chain == 'TRB', contig_fields = c('chain', 'v_gene','d_gene', 'j_gene', 'aa80'))

 ```