Improve Joint-RPCA

[TuomasBorman]

No description provided.

[TuomasBorman]

The old and new version are now producing the same results:

devtools::load_all()

data("ibdmdb")
mae <- ibdmdb
mae[[1]] <- transformAssay(mae[[1]], assay.type = "mgx", method = "rclr")
mae[[2]] <- transformAssay(mae[[2]], assay.type = "mtx", method = "rclr")

res <- getJointRPCA(mae, experiments = c(1, 2), assay.types = c("rclr", "rclr"), ncomponents = 3L, test.ratio = 10/60)

mats <- lapply(experiments(mae), function(x) assay(x, 1))
res2 <- .joint_rpca(mats, transform = c("rclr"))

res2$ord_res$samples <- res2$ord_res$samples[match(colnames(mae)[[1]], rownames(res2$ord_res$samples)), ]
all(abs(res - res2[["ord_res"]][["samples"]]) < 0.0001)

head(res2[["ord_res"]][["samples"]])
head(res)

res2$cv_stats
attributes(res)[["reconstruct_error"]]

[TuomasBorman]

Scripts to test that the Gemelli and our implementation is working similarly

test_joint_optspace.py
test_joint_optspace.txt (R file but it seems that GH does not support R files...)

[TuomasBorman]

Check this.

[TuomasBorman]

Check this.

[TuomasBorman]

Check this.

[TuomasBorman]

TODO:

• Add unit tests
• Tidy the example dataset. colData includes columns with names including many dots
• Add workflows

[TuomasBorman]

Check this.