mschubert / boundie

R package for differential expression on count data with parameter bounds

Hi there 👋

I'm Michael, a computational biologist. This means that I'm using computers and code to analyze biological data, often in collaboration with experimental scientists. I'm working on how cancer cells overcome their growth limitations and how we can use that to help patients. Here you can find some of the work I did, and more general purpose tools I wrote for anyone to use. Please reach out if you want to collaborate!

This is what I do for research 👨‍🔬 (full list 🔗)

	article	code
Developed a reliable method for estimating cell signaling pathways from gene expression		github bioc 📦
Showed how gene coexpression networks often reflect cell mixtures instead of regulation		github
Found a way how cancer cells can tolerate abnormal DNA content (aneuploidy, chromosomal instability) and a potential treatment, in collaboration with experimental scientists		transposon cgas_ko
Working on estimating DNA copy number of single-cell RNA sequencing	coming soon

Here are some of my open source contributions 🔠 (full list 🔗)

		status	code
clustermq	R package for efficient high performance computing		github cran 📦 testing ⚙️
narray	R package for simplifying array operations		github cran 📦
ebits	R bioinformatics toolkit incubator and data API		ebits data
	Software build scripts for the ArchLinux User Repository 🔗 and as Gentoo overlay 🔗		pkgbuilds overlay

boundie's Issues

Make sure pkg works with numeric 'x'

if just a vector and not matrix: ncol, rownames will not work

Track non-converged fits

And display warning with summary which genes failed to converge

Is trivial gradient solution accurate?

In gradient.r#L28, would it be better to solve for the betas to reach the closest point where mu>=0 instead of the trivial point mu=0?

Make fit work with non-integer counts

• suppressWarnings for poisson glm
• in nll, use function that is dnbinom if integer and interpolated otherwise

Can theta be negative?

maybe add constraint there