Forum:Bioinformatics commercial services as a risk factor for scientific misconduct

Question

[Angle45]

I have a question pertaining to the ethics and standard practices of scientific publication in the academic community, specifically with regards to bioinformatics. Today, I received an unsolicited email from a bioinformatics services company called Accurascience, claiming that I should consider using their commercial service for my bioinformatics analysis. I browsed through their website and could not find any mention or instruction or requirement for proper citation upon publication. I was also a bit taken aback by the following: http://www.scientificspam.net/?p=55

I decided to search "Accurascience" in PubMed, which returned zero results, and then followed up with a Google search to see whether or not people were actually citing Accurascience for their analysis. No relevant hits were found in a Google search either. I then read one of the clients' testimonials claiming that they had cited the company website in their PNAS paper. After finding this PNAS paper online, I noticed that no citation to Accurascience was actually given. Instead, the attribution of the analysis went to individual authors on the paper. -> Question: Is it really legal for people using a bioinformatics commercial service to attribute the bioinformatics work to themselves in publications? There just seems something majorly wrong with this.

I am not sure if this is a prime example of scientific misconduct or not, but I am confused as to why proper attribution wasn't given to Accurascience in the Hu et al. PNAS paper, whether it is ethical to outsource bioinformatics analysis like this (e.g., how many more ghost authorship papers like this exist?), and finally how it could possibly be ethical in the academic community to benefit from such commercial analysis without proper attribution to the company name that provides the service.

I appreciate any input you may have.

Answer 1

[pld]

Fee for service companies rarely get authorships and that's a normal practice. There are plenty of people who contribute that do not end up on a paper, which is reasonable. A paper isn't to be authored by anyone and everyone who took part or contributed in some way, else it would be filled with 100+ people. Normally the bioinformatics people would end up on the paper, but with a fee for service, it isn't unexpected that they aren't.

Lastly, it isn't unreasonable to say that the author reviewed and performed analysis. They very well may have gotten a pile of data back and the author, not the company, was the one who took the raw genome data and performed the analysis needed for the paper.

Lastly, it helps if you read the whole paper:

Selection of LTR Target Sites, WGS, Bioinformatics, and Statistical Analysis.

We used Jack Lin's CRISPR/Cas9 gRNA finder tool for initial identification of potential target sites within the LTR. Detailed WGS, bioinformatic, and statistical analyses are described in SI Materials and Methods.

And of course in the SI:

Demultiplexed read data from the sequenced libraries were sent to AccuraScience, LLC
(http://www.accurascience.com) for professional bioinformatics analysis.

Do your homework, especially if you're going to accuse someone of such a significant thing. Research misconduct is not taken lightly, don't accuse people of it without being absolutely sure.

I'd imagine that misconduct of this nature would be fairly tough to get by an editor like Anthony Fauci, the head of NIAID.

[Devon Ryan]

Nice job actually reading the paper rather than just skimming through it like the rest of us! This should really be the accepted answer.

[Angle45]

I am not sure if this is a prime example of scientific misconduct or no...

Definitely not accusing anyone here, read the post. Accusing someone of accusing however, I don't appreciate. I will, however, say that relegating Accurascience to the SI and misleading a whole group of scientists here (however much "skimming" we all really did) that the analysis was performed by the three authors (as stated in the main paper) is simply put: very bad practice. This should have been corrected by the eminent NIAID authority. It may have also been done with the intention to mislead: food for thought. Due credit should be given where due credit is due (especially in the main body of the paper), industry or academia alike. In this regard, I completely agree with Jean Karim-Heriche and Devon Ryan.

[pld]

That's not misleading, it isn't a bad practice, you didn't read the paper.

I didn't read the whole thing either, I went to materials and methods, saw that the NGS stuff was in a supplement and checked there. It took all of 30 seconds.

Fee for service companies or even core services/etc don't always end up on papers. They are otherwise acknowledged in the article as in this case (as are the methods used). You need to understand that there's a large difference between authorship and acknowledgment. Just because contributed doesn't mean they're an author and just because they're not listed as an author it doesn't mean that any dishonest or unethical actions have occurred.

Not to mention, everyone who contributes has a choice in if they're listed as an author. As others have mentioned, authorship isn't trivial, it has added responsibilities which translate to additional work. It is perfectly reasonable for a company to not want its employees as authors, it costs them money. 10 hours spent on writing, editing and aiding in reviewer response is 10 hours spent not working on other projects.

I don't see how there's any issue here.

[User 59]

Indeed. We've analysed thousands and thousands of exomes for customers, I can count the number of papers we're co-author on for that on the fingers of one hand. A few folk add an acknowledgement. 99.99% do neither.

[Devon Ryan]

I don't know, I probably would have worded all of that the same way they did it in the paper. Mentioning that in the SI is fine as far as I'm concerned. I don't think they were trying to mislead anyone.

[Angle45]

Devon, I'm saying that if the lion's share of the bioinformatics work was done by Accurascience, this should be made explicitly clear in the author contributions section and not be bluntly attributed to the paper authors. If not in contributions, then at least acknowledgements, at least so that it's not relegated to outside the main body of the paper, something not every reader looks at (especially on first reading of the paper). Otherwise, it's misleading.

[Devon Ryan]

Shoving stuff like this in the supplemental info section is pretty much par for the course these days. The exponential growth of supplemental information sections is a vastly bigger issue than any minor moot issue like this.

[pld]

This is especially true for high impact journals. It seems like most nature papers end up looking like abstracts compared to their SI.

[Damian Kao]

The bioinformatics analysis methods described in the supp. info doesn't seem better or worse than most publications out there. Whether that's acceptable is another issue.

I agree with Joe that I don't think we should discriminate against bioinformaticians just because they work in a non-academic setting.

Answer 2

[Devon Ryan]

Well, there's also the question of whether that paper actually used this service provider or if that's just total BS. You'd have to contact the authors and ask them. If they didn't use this provider, then accurascience is looking to get sued.

Let's just assume that the authors did outsource the service. The question then becomes what the authors' obligation is. The common practice is to mention in the methods that you outsourced some part of the analysis to a company. That then covers the authors somewhat if the analysis is later shown to be complete crap (N.B., I have yet to see a a single commercial provider that has produced anything other than complete crap analyses). What certainly can't be done is explicitly stating that the authors performed the analyses done by the commercial provider. However, that can be implied (it shouldn't be, but you're not going to get yourself in any real trouble by doing so). In that PNAS paper they mention that 3 of the authors performed data analysis. Since they didn't explicitly state that they provided all analyses they skirted the line. I'm not saying that this is right (it's absolutely not), but it's what's considered OK.

BTW, commercial service providers are almost never listed as co-authors unless that was part of the initial contract with them (there is then a discount for their services). You're paying them full price for a service, they have no authorship rights. A good example of how this works is with BGI. They have two-tiered pricing. You pay one amount if they're coauthors and another amount if they're not. Most providers don't offer any discount since they don't care about coauthorship (and wouldn't really want it anyway). Being a paper author comes with its own host of responsibilities that commercial providers don't normally want to deal with.

[Daniel]

I think there is a similarity to be had with black-box analysis pipelines too. There is a precedent for authors to write "we assessed diversity of the sample and found...." whereas I think it actually should read "Diversity was assessed using $pipeline which demonstrated...". There have been cases where I've seen FOSS have to push bug fixes because of actual incorrect data causing erroneous code being discovered, and if you said you did the analysis then you are responsible for that.

Answer 3

[Jean-Karim Heriche]

I think it is dishonest to take credit for work you haven't done. So if someone else, be it a company or a collaborator, analyses your data, they should at least be acknowledged if not listed as co-author. Also by not acknowledging other people's contributions, you're taking responsibility for their work i.e. if there's anything wrong, you'll take the blame, not them. Basically, there's nothing different between outsourcing to a company and outsourcing to the lab next door, the same ethical rules apply to both situations.

[Angle45]

Daniel Swan: "Indeed. We've analysed thousands and thousands of exomes for customers, I can count the number of papers we're co-author on for that on the fingers of one hand. A few folk add an acknowledgement. 99.99% do neither."

And then you wonder why bioinformaticians complain that they often get treated like technicians and non-equals in the laboratory. It's their own doing. Maybe if bioinformaticians changed their attitude and perceptions and (*gasp*) self-image to their own profession, perhaps the wet-lab community would start perceiving them a little differently. Ironic, eh?

[User 59]

Not at all. Do you think the people who do the DNA extraction and capture and library preparation and sequencing get on the paper? No, they don't. You don't seem to understand how a CRO or similar service operation works.

Answer 4

[Lars]

I can just add that I also got the AccuraScience email (spam). I have to admit, that I have never heard anything of any of their "Lead Bioinformaticians". So, what makes them the lead of all bioinformaticians, if not their work? Or are they the lead of all bioinformaticians working at companies? Well, if they are researcher, then their impact is measured by their publication track. Probably they are just the "Lead Bioinformaticians" of AccuraScience. :)

But, of course, it's a company and companies have to write like that. I directly put this mail in my trash, since it was too much. At least for me.

I know that this is not really related to the topic, but I just wanted to add that.

[pld]

I had never heard of you, so I guess you're terrible.

Lead doesn't mean "I publish a ton because I'm a scientific superstar". Imagine how many bioinformatics people employed by pharma companies or contracted by pharma that don't end up on papers or simply don't do work that ends up on papers. Industry doesn't publish like academia does and that doesn't make them better or worse as bioinformaticians. Lead probably means they're a mixture of management and scientist, you know, sort of like a PI.

Besides, if you'd do some googling you'd see that some (if not all) of them do publish:

• http://girke.bioinformatics.ucr.edu/
• http://www.researchgate.net/profile/Alberto_Riva2/publications
• http://jordan.biology.gatech.edu/page/software/index.html

What's with all the hate for industry? Not everyone wants to spend their lives dog fighting over minuscule chances for job security and funding.

[Istvan Albert]

I have also heard of some of the people on the team - and they are pretty good.

IMO in the future most of the data analytics that today is done by individuals will have to be performed in a more automated manner and at a scale and efficiency that only commercial companies will be able to deliver.

I suspect that the even today the quality of commercial data analysis is superior to that produced by "non-profit researchers". This latter comprises many thousands of people thrust into positions that they are not really qualified for.

[Angle45]

That's a bold conjecture, considering you run a Bioinformatics Consulting Center. You can't make a generalization that commercial is superior to academic without being unfair to academic researchers.

[pld]

I think it is a reasonable assumption. In an academic setting there's more margin for error and 'softer' impacts, that's the point of academic settings, to be trained. I also think you're more likely to find cases of where people are trying to do things outside their skill level/set.

In an industrial setting if you botch someone's data, you risk losing the customer and damaging your reputation. Reputation is everything and can be very hard to obtain/demonstrate even when you're doing a perfectly good job (as this thread shows). There's competition for customers and there's competition for jobs.

I doubt anyone would bat an eye if they heard a grad student or even a post doc screwed up, but if you heard that a company screwed up. You wouldn't bat an eye if you heard a tech was fired for screwing something up but a graduate student would have to really drop the ball to get the axe.

[Damian Kao]

Actually the "lead" in the job title is not supposed to signify any superiority in the entire field of bioinformatics. It's really just a company hierarchy title like "president" or "director". For example, software companies might have titles like "lead programmer" or "lead designer" signifying their seniority within the company.

Answer 5

[Alex Reynolds]

finally how it could possibly be ethical in the academic community to benefit from such commercial analysis without proper attribution to the company name that provides the service

I think it's even unethical for people to use or derive their tools or code from other's non-commercial software without proper attribution, but that has happened, too.

It's not pleasant to see your work end up rewritten into other people's projects without credit, but given the nature of open source it's hard to do much about it, especially if the parties who nick your work are high-profile.

I suspect commercial analysts find trading the lack of attribution for a bigger paycheck an acceptable bargain. I don't know that there is a solution for it for people in the open-source world. If you find a solution for enforcing attribution, let us know!

[Istvan Albert]

I think most reputable institutions do have proper avenues for dealing with these situations when they occur. First option would be to contact the authors and ask them to revise the paper, if that fails ask the journal and then escalate it if you think it is worth it.

Personally I think most situations like this can be explained by Hanlon's razor: "Never attribute to malice that which is adequately explained by stupidity." in that authors simply don't understand what should be cited when - and is not due to other motives.

http://en.wikipedia.org/wiki/Hanlon%27s_razor

[Alex Reynolds]

I don't want to name names or start a fight. It just bugs me when people make use of ideas without giving appropriate credit. It undermines the social compact that underlies non-profit research, when that happens.

[Angle45]

Ignorantia juris non excusat (Latin for "ignorance of the law excuses no one").

[pld]

I don't think Istvan was saying that they're innocent, just that there likely isn't intent to rob someone of their due credit.

[pld]

As far as I can tell, they referenced all of the tools they used correctly.