TODO.md

In jmw86069/genejam: Gene Jam

TODO for genejam

07sep2022

• Use case: Search for genes by description

• could involve something like pattern matching by grep

• For example, find "NF-KB" genes, or "NF-kB". Does it need to search for the Kappa symbol?

• Or search for "histone" or "ubiquitin".

• Consider changing default argument from intermediate="intermediate", to intermediate="ENTREZID".

• Consider a helper function for microarray annotation.

• Annotation of Affymetrix whole exome array probesets.

• Annotation of Illumina whole genome tiling array.

• Mimic the annotation step for oligo array classes.

• Consider allowing input SummarizedExperiment

• use either rownames(se) or rowData_colnames derived from colnames(rowData(se))

• process would extract the relevant data.frame from rowData(se), then store results back into the se object.
• freshenGenes_se(se, rowData_colnames=NULL) would use rownames(se)

• freshenGenes_se(se, rowData_colnames=c("AFFYID")) would use rowData(se)[, rowData_colnames]

• Could consider input types such as ExpressionSet, and the oligo array classes that are often hard to annotate easily.

30jun2021 (COMPLETE)

COMPLETE: Optional case-insensitive lookup

Current workflow requires two un-intuitive steps:

1. Prepare annotation library with get_anno_db(.., ignore.case=TRUE). This step also requires the full annotation library input, no convenience of providing only "SYMBOL2EG" for example. Also, this step is fairly slow - about 17 seconds to prepare three annotation environments. If not cached, obviously this step adds 17 seconds to every response time.

library(genejam)
try_dbs <- list(org.Hs.egSYMBOL2EG=org.Hs.egSYMBOL2EG,
   org.Hs.egACCNUM2EG=org.Hs.egACCNUM2EG,
   org.Hs.egALIAS2EGorg.Hs.egALIAS2EG=org.Hs.egALIAS2EG)
system.time(try_list <- lapply(try_dbs, function(i){
   get_anno_db(i, ignore.case=TRUE)
}))

1. Convert input to lowercase with tolower(), and query using the custom try_list object from step 1.

The solution seems to be the custom function genejam::imget() with case-insensitive mget(), with prefix i to indicate case-insensitive.

freshenGenes3(tolower(c("TRKB", "ApoE")), try_list=try_list)

COMPLETE: Fix bug when colname "intermediate" is supplied as input

Reproducible example (0.0.13.900 or less)

idf <- data.frame(Gene=c("MINA", "MINA", "GABRR3", "GABRR3"),
   intermediate=c("", "84864", "", "200959"))
idf
genejam::freshenGenes2(idf)

Output

    Gene intermediate intermediate_v1 SYMBOL                                                               GENENAME
1   MINA                        84864                                                                              
2   MINA        84864           84864  RIOX2                                                  ribosomal oxygenase 2
3 GABRR3                       200959                                                                              
4 GABRR3       200959          200959 GABRR3 gamma-aminobutyric acid type A receptor rho3 subunit (gene/pseudogene)

It creates a new column "intermediate_v1" then does not use it.

Desired outcome

Values should be left as-is in "intermediate" and new values should be used to fill in holes.

Ideally, user can specify the intermediate_colname so its values are recognized, potentially skipping the first step in freshenGenes() if there are no remaining colnames in the input data.

An argument intermediate_colname would make the process a bit more transparent.

Also, the colname intermediate_colname should not be split into multiple columns as is done with input data when it contains delimited values.

New workflow (version 0.0.14.900)

idf <- data.frame(Gene=c("MINA", "MINA", "GABRR3", "GABRR3", "APOE"),
   ENTREZID=c("", "84864", "", "200959", "84864"))
idf
genejam::freshenGenes2(idf, intermediate="ENTREZID")
genejam::freshenGenes2(idf, intermediate="ENTREZID", handle_multiple="first_hit")
genejam::freshenGenes2(idf, intermediate="ENTREZID", handle_multiple="first_try")
genejam::freshenGenes2(idf, intermediate="ENTREZID", handle_multiple="all")
genejam::freshenGenes2(idf, intermediate="ENTREZID", handle_multiple="best_each")
genejam::freshenGenes2(idf, intermediate="ENTREZID", include_source=TRUE)

Trickier example showing mix of input types:

idf <- data.frame(Gene=c("MINA", "RIOX2", "GABRR3", "GABRR3", "RIOX2,GABRR3", ""),
   ENTREZID=c("", "84864", "", "200959", "", "84864,200959"))
idf
genejam::freshenGenes2(idf, intermediate="ENTREZID")
genejam::freshenGenes2(idf, intermediate="ENTREZID", include_source=TRUE)

Final example showing input with only the intermediate column:

idf <- data.frame(Gene=c("MINA", "RIOX2", "GABRR3", "GABRR3", "RIOX2,GABRR3", ""),
   ENTREZID=c("", "84864", "", "200959", "", "84864,200959"))
idf
genejam::freshenGenes2(idf[,"ENTREZID",drop=FALSE], intermediate="ENTREZID")

Allow ENTREZID or EG as optional input column(s)

I bring this issue to the top on 30mar2021.

Problem space

It is not currently possible to provide ENTREZID values at input, since these values are expected to be in the "intermediate" column after the first step in the process. There is no annotation that takes ENTREZID and returns ENTREZID.

A close by-pass is to supply data with colname "intermediate", which works but breaks the actual use of "intermediate".

• Allow "ENTREZID" or "EG" as an input column, some mechanism to allow input data to have the intermediate value, without having to query an annotation.
• It would essentially mark one or more columns with something like "AsIs" and the values would be delimited and used to populate the "intermediate" column without query.
• There could be a special-case ann_lib="ASIS" that would allow values in the "AsIs" columns to pass through, which would give a mechanism to try querying another column first, then using the "ASIS" column as a secondary option.

Make searches case-insensitive

• This change requires something like better_mget() or perhaps more properly imget().
• The basic mechanism:

{r, imget} input <- c("Apoe", "h1-3", "Hist1H1C"); keys <- ls(org.Hs.egSYMBOL2EG) keymatch <- match(tolower(input), tolower(keys)) keysfound <- !is.na(keymatch); values <- as.list(jamba::nameVector(rep(NA, length(input)), input)) valuesfound <- mget(keys[keymatch[keysfound]], org.Hs.egSYMBOL2EG, ifnotfound=NA) values[keysfound] <- valuesfound; values;

After some testing, it appears the slow step is retrieving all the keys, which is also performed each time the data is queried, compounding the problem.

Current recommendation is to use get_anno_db(..., ignore.case=TRUE) which returns an environment where all keys are forced to lowercase. In this case, the environment consumes real memory, and takes longer to prepare. However once prepared, the environment can be passed directory to freshenGenes() for rapid operation.

Suggested mechanism

• asis_colnames=c("ENTREZID","EG") would define colnames c("ENTREZID","EG") as "ASIS" which means they are not to be queried.

• ann_lib=c("ASIS", "SYMBOL2EG", "ALIAS2EG") would process in order:

  1. take all values from the asis_colnames and delimit them to populate the "intermediate" column.
  2. query "SYMBOL2EG" using all columns except asis_colnames.
  3. query "ALIAS2EG" using all columns except asis_colnames.

• ann_lib=c("SYMBOL2EG", "ASIS", "ALIAS2EG") would process in order:

• query "SYMBOL2EG" using all columns except asis_colnames.

• Populate "intermediate" with delimited values from the asis_colnames using rules defined by handle_multiple. That is, when handle_multiple="first_try" then only rows with empty values in `"intermediate" column will be populated.
• query "ALIAS2EG" using all columns except asis_colnames.

jmw86069/genejam documentation built on Sept. 19, 2022, 1:53 p.m.