celltoprotein -- connecting Cell and Protein Ontologies

Vincent J. Carey, stvjc at channing.harvard.edu

March 25, 2026

Introduction

In a pair of papers from the Ventner Institute, Bakken et al. and Aevermann et al. discuss ontological implications of single-cell transcriptomics. A process of cell type definition via “necessary and sufficient marker gene” enumeration is introduced.

In this vignette we indicate how Cell Ontology, Relational Ontology, and Protein Ontology can be connected to assess formal relationships between declared cell types and plasma membrane features that can play a role in cell type definition.

Given a cell type, what proteins are noted as parts of its plasma membrane?

Connect to the relational ontology and search for CURIEs related to “plasma membrane”.

library(ontoProc2)
ro = semsql_connect(ontology="ro") 
search_labels(ro, "plasma membrane")

##      subject                           label
## 1 RO:0002104        has plasma membrane part
## 2 RO:0015015 has high plasma membrane amount
## 3 RO:0015016  has low plasma membrane amount

We have a helper resource for finding exact Cell Ontology names of cell types.

data("tag2cn", package="ontoProc2")
cd8reg = grep("CD8-positive.*regulatory", tag2cn, value=TRUE)
cd8reg

##                                                   CL:0000795 
##                 "CD8-positive, alpha-beta regulatory T cell" 
##                                                   CL:0000919 
##  "CD8-positive, CD25-positive, alpha-beta regulatory T cell" 
##                                                   CL:0000920 
##  "CD8-positive, CD28-negative, alpha-beta regulatory T cell" 
##                                                   CL:0001041 
## "CD8-positive, CXCR3-positive, alpha-beta regulatory T cell"

Now with these cell type identifiers, we can search for the proteins identified as “part of plasma membrane”. We need to use the CURIEs for precision

prtab = get_present_pmp(names(cd8reg))
library(DT)
datatable(prtab)

Given a protein, what cell types are asserted to possess it as a membrane part?

We pick two proteins and look for associated cell types.

prs = c("PR:000001094", "PR:000001380")
clk = cells_with_pmp(prs)
datatable(clk)

Some details

The “entailed edge” table of the Semantic SQL representation of Cell Ontology includes all assertions that are derivable from base axioms of the ontology.

cl = semsql_connect(ontology="cl")
cl

## <SemsqlConn>  prefix: CL  | labeled terms: 22,298

library(dplyr)
tbl(cl@con, "entailed_edge")

## # Source:   table<`entailed_edge`> [?? x 3]
## # Database: sqlite 3.51.2 [/Users/vincentcarey/Library/Caches/org.R-project.R/R/BiocFileCache/40e27456c620_cl.db]
##    subject        predicate       object        
##    <chr>          <chr>           <chr>         
##  1 UBERON:0001772 rdfs:subClassOf UBERON:0001772
##  2 UBERON:0019190 rdfs:subClassOf UBERON:0019190
##  3 GO:0051034     rdfs:subClassOf GO:0051033    
##  4 GO:0051033     rdfs:subClassOf GO:0051033    
##  5 GO:1904522     rdfs:subClassOf GO:1904522    
##  6 UBERON:0018685 rdfs:subClassOf UBERON:0018685
##  7 GO:0106027     rdfs:subClassOf GO:0106027    
##  8 GO:0050679     rdfs:subClassOf GO:0050679    
##  9 GO:1901647     rdfs:subClassOf GO:0050679    
## 10 GO:1904692     rdfs:subClassOf GO:0050679    
## # ℹ more rows

tbl(cl@con, "entailed_edge") |> count()

## # Source:   SQL [?? x 1]
## # Database: sqlite 3.51.2 [/Users/vincentcarey/Library/Caches/org.R-project.R/R/BiocFileCache/40e27456c620_cl.db]
##         n
##     <int>
## 1 2966269

We can look for statements that have “RO:0002104” as predicate:

tbl(cl@con, "entailed_edge") |> filter(predicate=="RO:0002104") |>
  as.data.frame() |> filter(grepl("PR:", object)) |> arrange(subject) |>
  datatable()