R - match combinations of nested list values from an index and return value

Question

Hi I have the two data sets. The first one is a list of genes linked to a given cluster (0-7):

# gene output

Cluster <- rep(0:7, each = 10)

Gene <- c("LMO3", "NEUROD6", "NFIB", "SNAP25", "RTN1", "CPE", "SOX11", "CSRP2", "VAMP2", "ID2", "EMX2", "LHX5-AS1","PEG10",
          "HES1", "TRH", "WLS", "TPBG", "RPS29", "CRABP2", "RSPO3", "RPL17", "RPL7", "PTMA", "RPL36A", "HMGN2", "H2AFZ",
          "NFIB", "PABPC1", "NEUROD6", "HNRNPH1", "PTN", "FABP7", "IGFBP2", "ID4", "C1orf61", "VIM", "RPS27L", "FABP5",
          "SDCBP", "BNIP3", "TCF7L2", "NEFL", "HMGCS1", "GAP43", "GPM6A", "SQLE", "ID4", "MSMO1", "SCOC", "BASP1", "TTR",
          "MEST", "TPBG", "MDK", "TMBIM6", "RCN1", "C8orf59","ID3","PKM", "PTN", "NCOR1", "ELAVL4", "NNAT", "ETFB",
          "STMN2", "TUBA1A", "GNG3", "MALAT1", "SOX4", "TUBB2B", "CRYAB", "GFAP", "CHCHD2", "HOPX", "LGALS1", "SCRG1", "ISG15",
          "AC090498.1", "B2M", "CLU")

df <- data.frame(cbind(Cluster, Gene))

The second is an index which provides cell-type annotations for specific combinations of genes:

# index

Type <- c("Radial Glia", "Excitatory Neuron ", "Inhibitory Neuron","Inhibitory Neuron",
          "IPC","Excitatory Neuron ","Radial Glia","Microglia","IPC","Inhibitory Neuron")

Subtype <- c("early", "Layer IV", "SST-MGE1", "SST-MGE1", "IPC-div2", 
             "Parietal and Temporal", "oRG/Astrocyte", "Microglia", "IPC-new", "MGE2")

Markers <- c("TOP2A AURK HMGB CTNNB1", "PPP1R1B SCN2A RORB CRYM", "DLX6-AS1 DLX1 SST DCX", "ERBB4 SST DLX2 DLX5 DLX6-AS1",
             "CCNB2 NEUROD4 KIF15 PENK HES6 ZFHX4 GLI3", "MEF2C STMN2 FLT ROBO CRYM", "AQP4 GFAP AGT DIO2 IL33",
             "C1QB AIF1 CCL4 C1QC", "CENPK EOMES", "CCK LHX6 SCGN SST")

index <- data.frame(cbind(Type, Subtype, Markers))

I am trying to find specific combinations outlined in Markers from the list of genes in my df. When such a match is found the corresponding type and subtype would be returned. However, there's a couple of caveats that I am finding very difficult to wrap my head around.

The lists for each cluster may contain multiple marker combinations - thus the function should go over each marker combination iteratively rather than stop when the first match is found.
The index-matching process should operate on each cluster separately - i.e. check genes in cluster 0 for marker matches and return type/subtype(s), then repeat the steps for cluster 1 etc.

My project data consists of dozens of df-like outputs made up of varied numbers of respective clusters, each containing hundreds to thousands of genes. I have tried my best to search solutions online but I am unfortunately drawing a total blank here.

Any help/toughts/suggestions would be greatly appreciated.

Edit:

the out put could look like so:

  Cluster    Gene        Type Subtype
1       0    LMO3 Radial Glia   early
2       0 NEUROD6            
3       0    NFIB            
4       0  SNAP25            
5       0    RTN1            
6       0     CPE

where a correct match(es) would add a row(s) to the df with corresponding type and subtype for each cluster, leaving the remainder empty (NAs).

user_7392175 · Answer

There is probably a much simpler way of doing this but here it is with a loop;

output = data.frame(Cluster=as.character(), Gene=as.character(), Type=as.character(), Subtype=as.character())

for(i in 1:nrow(df)){
  cluster = df[i,1]
  gene = df[i,2]
  type = index[grep(gene, index$Markers),]
  n_types = nrow(type)
  tmp = data.frame(Cluster=rep(cluster,n_types),
                   Gene=rep(gene, n_types), Type=type[,1], Subtype=type[,2])
  output = rbind(output,tmp)
}

user_4550695 · Answer

I’m assuming you want to annotate each cluster of genes with the types from the index, when all of the markers for a type are present in the cluster’s pool of genes.

I’m also going to use some simplified datasets; two simplified types in the index:

library(tidyverse)

index <- bind_rows(
  tibble(type = "AB", subtype = "X", markers = c("A", "B")),
  tibble(type = "BC", subtype = "Y", markers = c("B", "C")),
)

index
#> # A tibble: 4 x 3
#>   type  subtype markers
#>         
#> 1 AB    X       A      
#> 2 AB    X       B      
#> 3 BC    Y       B      
#> 4 BC    Y       C

And three different clusters that illustrate different matching scenarios:

clusters <- bind_rows(
  tibble(cluster = 0, genes = c("A", "B", "C")), # 2 matches
  tibble(cluster = 1, genes = c("B", "C", "D")), # 1 match
  tibble(cluster = 2, genes = c("C", "D", "E")), # No matches
)

clusters
#> # A tibble: 9 x 2
#>   cluster genes
#>      
#> 1       0 A    
#> 2       0 B    
#> 3       0 C    
#> 4       1 B    
#> 5       1 C    
#> 6       1 D    
#> 7       2 C    
#> 8       2 D    
#> 9       2 E

I would approach this by first making a function that returns matching types for a given pool of genes:

match_index <- function(genes) {
  matches <- index %>% 
    group_by(type, subtype) %>% 
    filter(all(markers %in% genes)) %>% 
    distinct(type, subtype)

  # If none matched, return a row of NAs  
  if (nrow(matches)) matches else matches[NA_integer_, ]
}

Then you can just summarise each cluster with the function:

clusters %>% 
  group_by(cluster) %>% 
  summarise(match_index(genes))
#> `summarise()` regrouping output by 'cluster' (override with `.groups` argument)
#> # A tibble: 4 x 3
#> # Groups:   cluster [3]
#>   cluster type  subtype
#>         
#> 1       0 AB    X      
#> 2       0 BC    Y      
#> 3       1 BC    Y      
#> 4       2

R - match combinations of nested list values from an index and return value

2 Answers