cole-trapnell-lab / garnett

Automated cell type classification
MIT License
99 stars 24 forks source link

cross species cell type transfer #33

Closed justinesjw closed 4 years ago

justinesjw commented 4 years ago

Hi,

Thank you for the Garnett it has been very useful.

Is it possible to train the classifier on mouse data and classify cells from human data?

Many thanks, Justine

hpliner commented 4 years ago

Hi Justine,

Glad it's been useful. We have had some success doing this. The main obstacle to trying it is the different gene names. The way I've done it before is make a "mouse-ified" human cds object or vice-versa (basically a mouse data cds with human gene names). Here's some code I've used in the past to generate a mouse-human gene table using biomaRt:

require("biomaRt")
human = useMart("ensembl", dataset = "hsapiens_gene_ensembl")
mouse = useMart("ensembl", dataset = "mmusculus_gene_ensembl")
genesV2 = getLDS(attributes = c("ensembl_gene_id"), 
                 filters = "ensembl_gene_id", 
                 values = fData(human_cds)$gene , 
                 mart = human, 
                 attributesL = c("ensembl_gene_id"), 
                 martL = mouse, 
                 uniqueRows=T)

Then a bunch of merges should get you a gene table and expression matrix with the right gene names that you can feed into new_cell_data_set.

Hope this helps, Hannah

justinesjw commented 4 years ago

Hi Hannah,

Thanks!

However, we realized that getLDS() do not convert all genes. We were able to convert more genes using this library. :) http://www.informatics.jax.org/downloads/reports/HMD_HumanPhenotype.rpt

Cheers! Justine

hpliner commented 4 years ago

Looks like this is resolved, reply or reopen if this continues to be an issue.