Updated clonotyping function based on implications for cells with different chain numbers than 1 VDJ 1 VJ chains.
This function offers two types of hierarchical clonotyping. The hierarchical option "single.chains" only merges cell with a single chain into clonotypes composed of cells with 1 VDJ 1 VJ chain. This is based on the assumption, that during mRNA capture and RT-PCR in GEMs, not all transcripts are captured and therefore cells may result missing a VDJ or VJ chain. The hierarchical option "double.and.single.chains" is based on the assumption, that cells with 1 VDJ and 2 VJ chains exist. For a review of the work concerning such cells as well as 2 VDJ 1 VJ cells please consult: https://doi.org/10.4049/jimmunol.1800904. The user may set a threshold of occurrence number above which cells with 1 VDJ 2 VJ chains are considered to be true and other cells with 1 VDJ 1 VJ, 1 VDJ 0 VJ and 0 VDJ 1 VDJ may be merged into the same clonotype by the strategy provided by the user. Cells with 2 VDJ chains are currently not considered in this process, as these are reported to be much rarer and, if appearing in the dataset are more likely to be doublets. We advice the user to carefully examine the output after hierarchical clonotyping before proceeding with further analysis. We thank Prof. Vijayanand as well as Vicente and Emmanuel from his lab for the discussions that have helped with improving the original Platypus clonotyping strategy.
VDJ_clonotype(
VDJ,
clone.strategy,
homology.threshold,
hierarchical,
triple.chain.count.threshold,
global.clonotype,
VDJ.VJ.1chain,
output.format,
platypus.version
)For platypus v2 output from VDJ_analyze function. This should be a list of clonotype dataframes, with each list element corresponding to a single VDJ repertoire. For platypus v3 VDJ output from the VDJ_GEX_matrix function (VDJ_GEX_matrix.output[[1]])
(Updated keywords, previous format is also functional) String describing the clonotyping strategy. Possible options are 10x.default, cdr3.nt, cdr3.aa, VDJJ.VJJ, VDJJ.VJJ.cdr3length, VDJJ.VJJ.cdr3length.cdr3.homology, VDJJ.VJJ.cdr3length.VDJcdr3.homology, cdr3.homology, VDJcdr3.homology. cdr3.aa will convert the default cell ranger clonotyping to amino acid based. 'VDJJ.VJJ' groups B cells with identical germline genes (V and J segments for both heavy chain and light chain. Those arguments including 'cdr3length' will group all sequences with identical VDJ and VJ CDR3 sequence lengths. Those arguments including 'cdr3.homology' will additionally impose a homology requirement for CDRH3 and CDRL3 sequences.'CDR3.homology',or 'CDRH3.homology' will group sequences based on homology only (either of the whole CDR3 sequence or of the VDJ CDR3 sequence respectively). All homology calculations are performed on the amino acid level.
Numeric value between 0 and 1 corresponding to the homology threshold forn the clone.strategy arguments that require a homology threshold. Default value is set to 70 percent sequence homology. For 70 percent homology, 0.3 should be supplied as input.
Character. Defaults to "none". This is an extention specifically for cells with aberrant numbers of chains (i.e. 0VDJ 1VJ, 1VDJ 0VJ, 0VDJ 2VJ, 2VDJ 0VJ). Cells with 2VDJ 2VJ are filtered out as these are most likely doublets. If set to "none" aberrant cells are assigned to their own clonotypes. If set to "single.chains" the function will proceed in two steps: 0. Prefiltering: cells with 2 VDJ 2 VJ chains as well as cells with 2 VDJ and any number of VJ chains are filtered out. 1. define clonotypes classically with all cells containing exactly 1VDJ 1VJ chains. 2. For cells with only a single chain (either VDJ or VJ), check if any clone exists, which matches the clonotyping criteria for this chain. If true, add this cell to that clone. If false, create a new clone containing that cell. In case that more than 1 existing clone matches the aberrant cell, the cell is assigned to the most frequent existing clone. Two reasons are behind this decision: 2.1. The aberrant cells is numerically more likely to be a part of the more frequent existing clone. 2.2 In case of a wrong assignment, the effect of the error is lower, if an already expanded clone is increase by one count, rather than a existing non-expanded clone being assigned a second entry and thereby resulting as expanded. Cells If set to "double.and.single.chains" the function will proceed as if set to "single.chains" but include two more steps 3. Check the frequency of each cell 1 VDJ 2 VJ chain exact clone (by exact nucleotide CDR3 matching). Only if this count exceeds the triple.chain.count.threshold, the clone is used as a "hub clone". This protects from merging clonotypes on the basis of rare doublets. 4. Merge existing clonotypes into the 1 VDJ 2 VJ clonotypes as they match with the assumption that e.g. a cell with 1 VDJ 1 VJ is part of that same clonotype, but missing a VJ chain due to stochastical sampling
Minimal occurrance frequency for any cell with more than 2 of either VDJ or VJ chain (e.g. 2 VDJ 1 VJ) for it to be considered as a trustworthy clone for hierarchical clonotyping ONLY when hierarchical is set to "double.and.single.chains". Defaults to 3, meaning that, an exact combination of three chains needs to appear in the dataset at least 3 times for it to be considered as a clone, into which other cells are merged. (For the counting of exact combination of chains CDR3 nucleotide string matching is used, even if clonotyping by homology)
Logical specifying whether clonotyping should occur across samples or only within a single sample (grouping via sample_id column).
Logical specifying whether cells other than once with 1 VDJ and 1 VJ chains should be considered.
Parameter output.format is deprecated. If non VGM-style output is required please refer to the function VDJ_clonotype. Output is VGM style VDJ by cell dataframe
Only "v3" available
Returns a VGM[[1]]-type dataframe. The columns clonotype_id and clonotype_frequency are updated with the new clonotyping strategy. They represent the "active strategy" that downstream functions will use. Furthermore extra columns are added with clonotyping information.New columns are named by clonotyping strategy so to allow for multiple clonotyping identifiers to be present in the same VDJ dataframe and make comparisons between these straighforward.
reclonotyped_vgm <- VDJ_clonotype(VDJ=Platypus::small_vgm[[1]],
clone.strategy="cdr3.nt",
hierarchical = "none", global.clonotype = TRUE)
#> Clonotyping strategy: cdr3.nt
#> Processing sample 1 of 1
#> Backing up 10x default clonotyping in columns clonotype_id_10x and clonotype_frequency_10x before updating clonotype_id and clonotype_frequency columns
reclonotyped_vgm <- VDJ_clonotype(VDJ=Platypus::small_vgm[[1]],
clone.strategy="cdr3.homology", homology.threshold = 0.5,
hierarchical = "single.chains", global.clonotype = TRUE)
#> Clonotyping strategy: cdr3.homology
#> Hierarchical clonotyping is specifically designed to incorporate cells with abberand numbers of chains. Filtering for 1VDJ 1VJ chain thereby defeats its purpose. Function will continue without filtering.
#> Filtered out 1 cells containing more than one VDJ AND VJ chain or two VDJ chains, as these likely correspond to doublets
#> Processing sample 1 of 1
#> Attempting to merge in 17 aberrant cells
#> Backing up 10x default clonotyping in columns clonotype_id_10x and clonotype_frequency_10x before updating clonotype_id and clonotype_frequency columns