Predict the group of neurons using instance or type information
Source:R/group.R
mcns_predict_group.RdPredict the group of neurons using instance or type information
Usage
mcns_predict_group(
ids,
method = c("auto", "fullauto", "group", "manc", "instance", "type", "pmanc", "all"),
badtypes = c(NA, "", "Lamina_R1-R6", "Descending", "KC", "ER", "LC", "PB",
"Ascending Interneuron", "Delta", "P1_L candidate", "LT", "MeMe", "PFGs", "Mi", "VT",
"ML", "EL", "FB", "Dm", "DNp", "FC", "OL", "T", "Y")
)Arguments
- ids
Body ids in any form understood by
mcns_ids. If you have a metadata dataframe as returned bymcns_neuprint_metathen this is ideal as that function is called under the hood.- method
A string specifying which of 5 methods to use to identify the group.
"all"means to return all 5, while"fullauto"means to look at each method in turn successively filling in missing group values. Method"auto"(the default) excludes predicted manc matches (see details).- badtypes
Values of the type column which should be ignored for the purposes of defining cell type groups. This will be because they contain bad values or because the types are too broad to be very useful.
Value
For method="all" a dataframe as returned by
mcns_neuprint_meta with additional columns
instance_group and type_group. Otherwise a numeric vector.
Details
Grouping information for neurons in the male cns is presently scattered in several locations. These include the numeric group field, the type field or the instance field. If the type field has the same value, the neurons should form a group. However there are some values that are known to be bad and these are excluded.
An additional source of group information comes from matches of VNC neurons
to the MANC dataset. These either come as curated matches (where the
manc_group column has been entered in Clio, method="manc") or
as predicted matches (based on the manc_bodyid column,
method="pmanc").
method="pmanc" should be used with caution since a significant
percentage of these matches are wrong. However, since the majority should
be correct, they may still be a useful source of group information e.g. for
connectivity clustering which is typically not that sensitive to errors.
Given this situation method='auto' (the default) only uses curated
matches (method="manc"). Select method='fullauto' to use the
predicted MANC matches as a fall-back.
Examples
# \donttest{
library(dplyr)
# return all body ids with a group type or instance
tig_ids=mcns_ids('where:exists(n.group) OR exists(n.type) OR exists (n.instance)')
allg=mcns_predict_group(tig_ids, method = 'all')
# neurons where the recorded group and instance group disagree
allg %>% filter(!is.na(group) & !is.na(instance_group) & group!=instance_group)
#> bodyid post pre downstream upstream synweight flywireType group
#> 1 28195 661 133 987 661 1648 CB1761,CB2741 28195
#> 2 52953 751 137 981 751 1732 CB1761,CB2741 28195
#> 3 129990 759 147 1150 759 1909 CB1761,CB2741 28195
#> 4 41250 870 158 1125 870 1995 CB1761,CB2741 28195
#> 5 46029 736 151 1053 736 1789 CB1761,CB2741 28195
#> 6 48918 787 434 3227 787 4014 CB1159,CB2015,CB2669 48918
#> 7 49406 741 471 3117 741 3858 CB1159,CB2015,CB2669 48918
#> hemibrainType name itoleeHl somaSide statusLabel superclass
#> 1 <NA> 28915_L LALa1_anterior L Roughly traced cb_intrinsic
#> 2 <NA> 28915_L LALa1_anterior L Roughly traced cb_intrinsic
#> 3 <NA> 28915_L LALa1_anterior L Roughly traced cb_intrinsic
#> 4 <NA> 28915_R LALa1_anterior R Roughly traced cb_intrinsic
#> 5 <NA> 28915_R LALa1_anterior R Roughly traced cb_intrinsic
#> 6 <NA> 48919_L LHp2 L Roughly traced cb_intrinsic
#> 7 <NA> 48919_R LHp2 R Roughly traced cb_intrinsic
#> supertype type class entryNerve fruDsx matchingNotes rootSide assignedOlHex1
#> 1 23508 <NA> <NA> <NA> <NA> <NA> <NA> NA
#> 2 23508 <NA> <NA> <NA> <NA> <NA> <NA> NA
#> 3 23508 <NA> <NA> <NA> <NA> <NA> <NA> NA
#> 4 23508 <NA> <NA> <NA> <NA> <NA> <NA> NA
#> 5 23508 <NA> <NA> <NA> <NA> <NA> <NA> NA
#> 6 18606 <NA> <NA> <NA> <NA> <NA> <NA> NA
#> 7 18606 <NA> <NA> <NA> <NA> <NA> <NA> NA
#> assignedOlHex2 mancBodyid mancGroup mancType somaNeuromere subclass trumanHl
#> 1 NA NA NA <NA> <NA> <NA> <NA>
#> 2 NA NA NA <NA> <NA> <NA> <NA>
#> 3 NA NA NA <NA> <NA> <NA> <NA>
#> 4 NA NA NA <NA> <NA> <NA> <NA>
#> 5 NA NA NA <NA> <NA> <NA> <NA>
#> 6 NA NA NA <NA> <NA> <NA> <NA>
#> 7 NA NA NA <NA> <NA> <NA> <NA>
#> dimorphism synonyms birthtime mancSerial mcnsSerial exitNerve serialMotif
#> 1 <NA> <NA> <NA> NA NA <NA> <NA>
#> 2 <NA> <NA> <NA> NA NA <NA> <NA>
#> 3 <NA> <NA> <NA> NA NA <NA> <NA>
#> 4 <NA> <NA> <NA> NA NA <NA> <NA>
#> 5 <NA> <NA> <NA> NA NA <NA> <NA>
#> 6 <NA> <NA> <NA> NA NA <NA> <NA>
#> 7 <NA> <NA> <NA> NA NA <NA> <NA>
#> receptorType somaLocation tosomaLocation status totalNtPredictions
#> 1 <NA> 66620,24888,15730 Traced 133
#> 2 <NA> 65254,26033,14863 Traced 137
#> 3 <NA> 65850,24844,15111 Traced 147
#> 4 <NA> 30730,36422,22418 Traced 158
#> 5 <NA> 30920,37020,23082 Traced 151
#> 6 <NA> 68976,14832,32846 Traced 434
#> 7 <NA> 28544,10466,32448 Traced 471
#> predictedNtConfidence predictedNt celltypeTotalNtPredictions
#> 1 0.8200724 gaba 133
#> 2 0.8341187 gaba 137
#> 3 0.7992291 gaba 147
#> 4 0.8392468 gaba 158
#> 5 0.7578290 gaba 151
#> 6 0.8384565 acetylcholine 434
#> 7 0.6560721 acetylcholine 471
#> celltypePredictedNt celltypePredictedNtConfidence consensusNt voxels
#> 1 unclear NA gaba 229963626
#> 2 unclear NA gaba 248926814
#> 3 unclear NA gaba 241019652
#> 4 unclear NA gaba 297966845
#> 5 unclear NA gaba 271122994
#> 6 unclear NA acetylcholine 457493970
#> 7 unclear NA acetylcholine 485994463
#> soma instance_group type_group mancGroup_group pmanc_group
#> 1 TRUE 28915 NA NA 28195
#> 2 TRUE 28915 NA NA 28195
#> 3 TRUE 28915 NA NA 28195
#> 4 TRUE 28915 NA NA 28195
#> 5 TRUE 28915 NA NA 28195
#> 6 TRUE 48919 NA NA 48918
#> 7 TRUE 48919 NA NA 48918
# }
if (FALSE) { # \dontrun{
# neurons where the recorded group and type group disagree
type_group_mismatch <- allg %>% filter(!is.na(group) & !is.na(type_group) & group!=type_group)
allg %>%
filter(group %in% type_group_mismatch$group | type_group %in% type_group_mismatch$type_group) %>%
select(bodyid, type, name, group, type_group, instance_group) %>%
arrange(type, group) %>% View
} # }