License

[drisso]

In #Venice-2026-spatial > Question on SpatialExperiment licensing @rcastelo asked if it's possible to change the SingleCellExperiment license to something less restrictive than GPL-3, e.g., Artistic-2.0.

I don't recall the specific reason we chose GPL-3, but I'm in favor of this change since SummarizedExperiment (and much of Bioconductor) is licensed under Artistic-2.0.

@LTLA what are your thoughts?

Tagging @vjcitn too in case he has some input.

I believe that all authors need to agree to change the license (see e.g., here), so starting this discussion to see first if @LTLA would be OK with this.

Cheers,
Davide

[LTLA]

I don't know if I originally chose the license, but if I did, it's because I was (and still am) a viral copyleft sympathizer. So I would be mildly saddened if that was changed.

I didn't think there were any problems with using GPL R packages within commercial environments. I vaguely remember being given the all-clear to use them within Genentech when I was employed there... something about linking. In any case, R itself is licensed under GPL-2/3, so a reasonable compromise would be to add GPL-2 | GPL-3 to SingleCellExperiment's License:, given that you wouldn't be able to use this package without satisfying R's own license conditions.

All that said, I guess I'm not that much of a true believer, given that I use the MIT license for most of my own packages. Mostly because Genentech Legal forced me to do so when I was publishing software from there, and I haven't bothered to change my default set-up because GPL-3's LICENSE file is so damn long that it's a pain to grab a copy.

[drisso]

I'm really not an expert on licenses, but reading a bit about R's own license and Hadley Wickham & Jenny Bryan's take on it it seems that if you simply use SingleCellExperiment's exported functions you are not bound by the "viral" aspect of GPL.

In light of this, I'm tempted to say that things are fine as they are, but I'm happy to hear other opinions.

[rcastelo]

Hi thanks for opening this issue, aside from how much R and GPL-licensed packages can be used in commercial environments, all upstream Bioconductor dependencies from SingleCellExperiment (S4Vectors, GenomicRanges, SummarizedExperiment and BiocGenerics) are licensed with Artistic-2.0.

Would not make sense that being SingleCellExperiment one of the current core packages of Bioconductor, would be also licensed with Artistic-2.0?

This is the code snipped I used to find out this information, there may be easier ways to do it:

library(BiocPkgTools)

depdf <- buildPkgDependencyDataFrame(repo=c("BioCsoft", "CRAN"),
                                     dependencies=c("Depends", "Imports"))
scedeps <- pkgDepMetrics("SingleCellExperiment", depdf)
for (d in rownames(scedeps)) 
    scedeps[d, "License"] <- packageDescription(d, fields="License")
scedeps                                                                                                
                     ImportedAndUsed Exported Usage DepOverlap DepGainIfExcluded         License         
S4Vectors                          2      291  0.69       0.28                 0    Artistic-2.0         
GenomicRanges                      2      108  1.85       0.40                 0    Artistic-2.0         
DelayedArray                       3      149  2.01       0.80                 0    Artistic-2.0         
SummarizedExperiment               2       86  2.33       0.96                 2    Artistic-2.0         
BiocGenerics                       7      133  5.26       0.20                 0    Artistic-2.0         
stats                             NA      454    NA       0.04                 0 Part of R 4.5.2         
utils                             NA      228    NA       0.04                 0 Part of R 4.5.2         

[LTLA]

To be clear, "core" in this context would mean "maintained by the BioC-core team". SingleCellExperiment isn't a core package in this respect, even though it lies at the base of the stack for many BioC-based workflows.

I'm not convinced that consistency is a good argument for changing the license. I'm not even consistent with my own licenses across projects - sometimes by necessity due to vendored dependencies, sometimes because I just forget what I used in a previous project, and other times because I just like to mix it up for some variety. As long as it's an open-source license, it seems fine to let people choose what they want.

If commercial use really is a concern, I'm willing to change it to GPL-2 | GPL-3, in which case SingleCellExperiment would be no less restrictive than R itself.

[rcastelo]

I meant "core" in the sense of importance, as you say, SingleCellExperiment lies at the base of the stack for many BioC-based workflows. As I see it, SingleCellExperiment defines the "core" data structure for single-cell data in Bioconductor, the one that package reviews advocate to use when new packages on single-cell analysis are submitted, currently 43rd most downloaded package. I do see license consistency as an important feature for this kind of upstream packages that define object classes that interoperate between them. The issue that triggered this question is the fact that lawyers at companies may ban developers to deploy GPL-licensed Bioconductor packages in their stack, which potentially reduces the adoption of Bioconductor software by companies, some of which may be contributing the technology that produces the data for which we develop methods.

[LTLA]

The issue that triggered this question is the fact that lawyers at companies may ban developers to deploy GPL-licensed Bioconductor packages in their stack

That seems unlikely. Having worked in a company, and having known people working in other companies, I am yet to hear any concerns about commercial use of GPL software. IANAL but the GPL doesn't prohibit commercial use, or even distinguish between different categories of users; it seems to be all about source code availability + modification + redistribution.

Perhaps your comment would be more relevant if people are planning to redistribute SingleCellExperiment in some application. But if you are going to worry about that, then there's a much deeper stack to worry about (R itself, Linux, etc.).

Anyway, if you want consistency, then making all packages GPL-2 | GPL-3 would seem to be most appropriate, given that R itself is released under those licenses.

[rcastelo]

I see, I have no experience working in a company as you do, so thanks for giving advice on how to navigate the software licensing landscape from a corporate perspective. I guess then that having SingleCellExperiment licensed with GPL2 | GPL-3 should do the job. In any case let me invite @patrickjdanaher to the conversation, in case he has any additional question, because it was him who raised this issue within the context of the hackathon on spatial omics in which we participated last week.

Patrick works at Bruker Spatial Biology, where they analyze spatial omics data (see here), but were not aware of the Bioconductor ecosystem and discussing on using it (particularly the SpatialExperiment package and upstream dependencies) and possibly integrating it into their software, this licensing discussion popped up.

[HelenaLC]

Random note that Seurat has an MIT license. If Aaron thinks GPL should be fine, I believe him and wouldn't know any better... Just thinking that having commercial vendors using/developing for Bioc has its upsides... see 10x-Seurat. So anything we can do to "allow" that "reasonably" would make sense to me (but I clearly don't now the legal detail nor inner company working :).

[rcastelo]

Good point, I did not think about comparing, but now that you mention it, scverse Python libraries seem to use also non-copyleft licenses such as BSD-3. Maybe because I'm not an expert on these matters either, I'd think that reflecting into what other projects do leaving sympathies aside might illuminate this kind of decision, but the authors should decide.

[rcastelo]

Hi, in this message in Zulip, Patrick has confirmed that the lawyers at Bruker finally approved using OSS with GPL-2 | GPL-3 licenses. So, it would be great if all authors would agree on switching the SingleCellExperiment license from GPL-3 to GPL-2 | GPL-3. Thanks!