TODO Development Tasks

[jmitchell81]

Overview

Task list for development priorities for domino

Intended Outcome

Serves as a central resource for highlighting developments or bugs to address and sharing them with the team who will track when these tasks are accomplished.

Task List

High priority

• Include a differential expression method within domino to prioiritize interactions based on ligand expression
• Disregard complex interactions when using gene ortholog conversions
• Add example usage to all function documentation
• build_domino fails when no TFs have a p-value below the threshold for one or more clusters.

Priority

• Linkage based on regression analysis
• Edit gene conversion to use first returned ortholog
• Replace counts data with normalized expression
• Ensure Domino classes are compliant with S4 standards https://adv-r.hadley.nz/s4.html
• Include helper function for reformatting CellTalkDB mouse or human databases http://tcm.zju.edu.cn/celltalkdb/
• If using a Seurat object as input data, add dependent of the Seurat package to read the active Assay
• Clarify the formatting of the tf.activity matrix to match gene x cell formatting used in R as opposed to Python cell x gene
• test outputs and plotting when running Domino without clustering
• The help text for build_domino is incomplete, the description for the first 2 options is missing
• Clarify that counts matrix must have gene names as row names for for expression checks
• Update references to domino package name to Domino2

Plotting

CI/CD & maintenance

• Github workflow for build, check, BiocCheck for quality control #46

Complete

• Allow alternative biomaRt mirrors for gene ortholog conversions c96a9ab
• per-cluster checks of receptor expression 82d2a97
• dedicated function for cellphoneDB reference processing 971d146
• Fix hard-coded receptor expression threshold for remove_rec_dropout in create_domino() 971d146
• AnyMatrix for class slots replaced by dgCMatrix class 518b74a
• create_rl_map_cellphonedb function is not found after the github install of domino 518b74a
• Ensure transcription factor targets are tracked in the domino object Issues running with current version of pySCENIC Elisseeff-Lab/domino#16 (comment) 9bbcd4a
• Move help text for gene_conv from create_domino to create_rl_map_cellphonedb 9474d91
• Include a check for the ellipses in the auc matrix that warns and removes the ellipses for compatibility with regular searching 679cf66
• Include Matrix package as a dependency to use rowSums() and allow dcGMatrix sparse formats 518b74a
• Clarify variable names "features" and "df" to "tf.activity", "regulons" 9474d91
• Aggregate expression of genes encoding ligand complexes in the cl_signaling matrices fdda20f
• Correct the setting the correlation of receptors in regulons to 0 679cf66

[kjkrishnan]

A few additional things I've noticed as I put together vignettes are below. I have been using the PBMC3k data that @jmitchell81 included in the vignette with 25 TFs per cluster and 25 receptors per TF, correlation threshold 0.25 and other parameters default (in case that matters for troubleshooting).

• gene_network plot title is not working for more than one cluster in the clust argument with no OutgoingSignalingClust

gene_network(dom, levels(dom@clusters))

results in the below output:

---

• gene_network plot giving subscript out of bounds error if vectors of clusters are passed to clust and OutgoingSignalingClust. For example:

gene_network(dom, levels(dom@clusters), OutgoingSignalingClust = "CD16_monocyte")

outputs an error:

Error in dom@cl_signaling_matrices[[cl]][, OutgoingSignalingClust] :
subscript out of bounds

---

• signaling_network gives an error requiring finite 'xlim' values when using the showIncomingSignalingClusts argument. For example, if I wanted to look at incoming signal to the CD8 T cell clusters:

signaling_network(dom, showIncomingSignalingClusts = "CD8_T_cell")

OR

signaling_network(dom, clusts = levels(dom@clusters), showIncomingSignalingClusts = "CD8_T_cell")

give an error:

Error in plot.window(...) : need finite 'xlim' values

---

• circos_ligand_receptor is sometimes throwing an error at the step where expression values are obtained from cl_signaling matrices (ex. LILRB2). However, this is inconsistent. For example,

circos_ligand_receptor(dom, receptor = "CD74", cell_colors = cols)

gives output:

but

circos_ligand_receptor(dom, receptor = "LILRB2", cell_colors = cols)

gives error:

Error in data.frame(origin = paste0(cell_names, "-", l), destination = receptor, :
arguments imply differing number of rows: 9, 1, 0

---

I am not sure to what extent these are usage issues on my end (in which case, we should adjust the function documentation to be more precise about what is or isn't allowed, and maybe include more informative error messages as well), but these are all examples of functionality I would expect to work (ie, I wasn't trying to break anything on purpose), so I figure we should look into that. @snag-gh let me know if you need more information or reproducible examples!