I was wondering, and I'm sorry if this is outside of the scope, but do people use negative sample controls in their sequencing? Like I'm running several Salmonella genomes, should there be just a barcode that contains "nothing". I have been asking the oracle (Google) on how to properly set up a sequencing run/experiment with little guidance. I know that the kits for the different platforms contain positive control sequences for run analytics and calibration but the different trainings have never said to include a negative reaction.
If this was PCR I would have a various negative controls, mock extraction, non template, no RT, etc.. I understand the cost of sequencing was, at one time, prohibitive of a lot of controls in the reactions but now it's nothing to add a blank to a rxn for a core and a minimal cost for a small lab.
Also, if you do run a negative control. How do you use it? You can use it to see if there is contamination but is it useful? Would you use the contamination found in the negative control to map to first to subtract out any other contamination?
Contamination is rampant in sequencing, and it's particularly hard to detect when sequencing human-related bacteria like Salmonella. I highly recommend running at least one control. If you get contamination that is taxonomically distant from Salmonella, you're in luck - it will be fairly easy to remove digitally.
Note that also, there is cross-contamination caused by barcode misassignment, misreading, chimerism, and various other effects; controls are useful to measure that, particularly in sensitive experiments like single-cell sequencing.
There are steps in the Illumina library prep. where you create the library without DNA, just like you would for PCR (although obviously it depends if the sequencing folks are doing it or not). This is to make sure that your library buffers do not contain contaminating DNA - or only the DNA expected, such as adapters, barcodes, etc - for all the same reasons as for PCR. PCR is generally one of the steps in sequencing anyway.
But what you seem to be suggesting is loading nothing but adapters/barcodes onto the flowcell to see if something gets sequenced regardless. Perhaps, for example, there is already DNA inside the flowcell, for example.
Well, there are two practical issues with this idea. The first is that while we can sequence things very cheaply thanks to multiplexing/barcoding, a negative control like the one you describe would need to happen to take up at least one whole lane of a flowcell, which is still a lot of money. It would also probably cause the sequencing machine to raise an error and abort. But most of all, such a control would only tell you is that the lane was clean - but no assumptions can be made for the other lanes with actual library DNA in it. Unlike PCR buffers, you can't reuse a flowcell (well... you can, you're just not legally allowed to).
Perhaps the machine itself does a pre-experiment diagnostic check to make sure the flowcell is clean of contaminants and the optical sensor is not registering anything, but unless that is i the Illumina user guide (which is outside the scope of Biostars), then its likely to be a trade secret as to how exactly such controls are performed.
They way you described the idea of a negative control would be terribly expensive and not useful. So perhaps this is why it's not done. Do you know if people use mock sample controls? Like from extraction forward? Again it would end up taking the place of a sample.
I guess they are really not needed, it just feels weird never doing one after all the other experiments that require them.
You're welcome - and by the way although I was negative, I think you raise a good point. We are trusting Illumina here that they're not making any mistakes re: contaminated DNA in the machine before you load your sample.
However, and this is a little off-topic but I only mention it to try to answer your last sentence, old-school experiments and old-school thinking are becoming less and less common in science - for better or for worse. 200 years ago, a theory was usually "proven" when someone invented a single conclusive, usually elaborate, clever and well-designed experiment who's only explanation is that the hypothesis in question is true. Usually in an auditorium in front of pompous/"enlightened" gentry. From electromagnetic induction to vaccines, there are many of these gold-standard experiments in the text books. Modern research, particularly all the next-gen sequencing assays, are a bit different. I think the philosophy of young scientists these days are more inline with the tech-startups around them - deliver early and often, and make improvements as you go. Perhaps this is due to the pressure of modern academia more than anything else. Anywhoo - my point is you can have all the negative controls you like, but the only control that counts for anything these days is when someone else on the other side of the world reproduces and reaffirms your results/conclusions. I have seen some absolutely nonsensical and boardline ludicrous experimental designs in the last 4 years -- but it's OK because we've replace mistrusting our equipment with mistrusting each other.... and there's arguments to suggest that was for the best.
The downside of usurping the sanctity of experimental design is that we're now in a place where people can literally make shit up and it doesn't really matter, so long as other people can reproduce it. I find that grossly troubling, particularly as NGS summary results (BigWigs, etc) are close to impossible to examine forensically. But my point is simply that negative controls are no longer in the minds of researchers, as there are bigger things to worry about :)
We are trusting Illumina here that they're not making any mistakes re:
contaminated DNA in the machine before you load your sample.
I wouldn't make that assumption; there have been issues with library carryover in the past (e.g., see this thread). But it's only an issue for certain types of analyses, such as tumor heterogeneity.
I'm hard to read because i'm not sure myself :) Of course I'd want to run the best analysis possible, but optimizing for time and money is so important these days that if we choose to blindly ignore those factors and proceed with "the perfect experiment" then we will probably get scooped and/or run out of money. And it will be frustrating as hell because you can't debug a commercial kit with unknown reagents. I'm frustrated, but it's not sarcasm. Science is just changing, for better or worse. 5 years ago I used to hear people say all the time "never use pre-cast gels" because you can't be sure of them unless you made them yourself. These days you're more likely to hear "why don't you stop messing around and just use a pre-cast gel?". Done is better than perfect :)
Entirely off topic, but pre-cast gels are awesome :p But I agree on the commercial kit with unknown reagents, that annoys me when I try to adapt or develop something new
We typically include negative controls from the extraction and amplification steps for our metagenomics work. They get separate barcodes and are quantified and included in the sequencing. We currently just look to (hopefully) see that there are few reads corresponding to these barcodes.
Most of the negative controls that you describe are designed to detect contamination or non-specific signals. If you're concerned about those issues for library prep, it's simple enough to include a no-input/mock sample control. It's not necessary to sequence it, however. Just like PCR, your expectation is no product - only if you detect product (e.g., by Bioanalyzer or qPCR) is there a potential source of error. At that point, it might be worthwhile to obtain a few hundred reads (by adding it to your sample pool at low concentration) to determine if the data will adversely affect your analysis or not. But it's not necessary to dedicate an entire lane for sequencing.
We perform routinely multiplex amplicon sequencing (on MiSeq) and do incorporate negative controls, which get barcoded and are carried through the entire process exactly like real samples. Since we sequence hundreds up to 1536 samples, a few negative controls are no problem.
I'm the case of amplicon sequencing you can figure out which amplicon is present and as such find which part of the assay is affected. Another thing is figuring out which step of your protocol resulted in the contamination, by checking when negative controls were negative (for example run an agarose gell), and when those became positive.
Contamination is rampant in sequencing, and it's particularly hard to detect when sequencing human-related bacteria like Salmonella. I highly recommend running at least one control. If you get contamination that is taxonomically distant from Salmonella, you're in luck - it will be fairly easy to remove digitally.
Note that also, there is cross-contamination caused by barcode misassignment, misreading, chimerism, and various other effects; controls are useful to measure that, particularly in sensitive experiments like single-cell sequencing.
There are steps in the Illumina library prep. where you create the library without DNA, just like you would for PCR (although obviously it depends if the sequencing folks are doing it or not). This is to make sure that your library buffers do not contain contaminating DNA - or only the DNA expected, such as adapters, barcodes, etc - for all the same reasons as for PCR. PCR is generally one of the steps in sequencing anyway.
But what you seem to be suggesting is loading nothing but adapters/barcodes onto the flowcell to see if something gets sequenced regardless. Perhaps, for example, there is already DNA inside the flowcell, for example.
Well, there are two practical issues with this idea. The first is that while we can sequence things very cheaply thanks to multiplexing/barcoding, a negative control like the one you describe would need to happen to take up at least one whole lane of a flowcell, which is still a lot of money. It would also probably cause the sequencing machine to raise an error and abort. But most of all, such a control would only tell you is that the lane was clean - but no assumptions can be made for the other lanes with actual library DNA in it. Unlike PCR buffers, you can't reuse a flowcell (well... you can, you're just not legally allowed to).
Perhaps the machine itself does a pre-experiment diagnostic check to make sure the flowcell is clean of contaminants and the optical sensor is not registering anything, but unless that is i the Illumina user guide (which is outside the scope of Biostars), then its likely to be a trade secret as to how exactly such controls are performed.
Thank you for the feedback John,
They way you described the idea of a negative control would be terribly expensive and not useful. So perhaps this is why it's not done. Do you know if people use mock sample controls? Like from extraction forward? Again it would end up taking the place of a sample.
I guess they are really not needed, it just feels weird never doing one after all the other experiments that require them.
You're welcome - and by the way although I was negative, I think you raise a good point. We are trusting Illumina here that they're not making any mistakes re: contaminated DNA in the machine before you load your sample.
However, and this is a little off-topic but I only mention it to try to answer your last sentence, old-school experiments and old-school thinking are becoming less and less common in science - for better or for worse. 200 years ago, a theory was usually "proven" when someone invented a single conclusive, usually elaborate, clever and well-designed experiment who's only explanation is that the hypothesis in question is true. Usually in an auditorium in front of pompous/"enlightened" gentry. From electromagnetic induction to vaccines, there are many of these gold-standard experiments in the text books. Modern research, particularly all the next-gen sequencing assays, are a bit different. I think the philosophy of young scientists these days are more inline with the tech-startups around them - deliver early and often, and make improvements as you go. Perhaps this is due to the pressure of modern academia more than anything else. Anywhoo - my point is you can have all the negative controls you like, but the only control that counts for anything these days is when someone else on the other side of the world reproduces and reaffirms your results/conclusions. I have seen some absolutely nonsensical and boardline ludicrous experimental designs in the last 4 years -- but it's OK because we've replace mistrusting our equipment with mistrusting each other.... and there's arguments to suggest that was for the best.
The downside of usurping the sanctity of experimental design is that we're now in a place where people can literally make shit up and it doesn't really matter, so long as other people can reproduce it. I find that grossly troubling, particularly as NGS summary results (BigWigs, etc) are close to impossible to examine forensically. But my point is simply that negative controls are no longer in the minds of researchers, as there are bigger things to worry about :)
I wouldn't make that assumption; there have been issues with library carryover in the past (e.g., see this thread). But it's only an issue for certain types of analyses, such as tumor heterogeneity.
Wow 0.2% is quite substantial! Thanks for this harold :)
@John - It's hard to tell whether you're pro- or anti- "gold-standard experiments".
Are you saying that well-designed experiments are no longer a priority, or is that angry sarcasm?
If negative controls are no longer in the minds of researchers, then they bloody well should be.
I'm hard to read because i'm not sure myself :) Of course I'd want to run the best analysis possible, but optimizing for time and money is so important these days that if we choose to blindly ignore those factors and proceed with "the perfect experiment" then we will probably get scooped and/or run out of money. And it will be frustrating as hell because you can't debug a commercial kit with unknown reagents. I'm frustrated, but it's not sarcasm. Science is just changing, for better or worse. 5 years ago I used to hear people say all the time "never use pre-cast gels" because you can't be sure of them unless you made them yourself. These days you're more likely to hear "why don't you stop messing around and just use a pre-cast gel?". Done is better than perfect :)
Entirely off topic, but pre-cast gels are awesome :p But I agree on the commercial kit with unknown reagents, that annoys me when I try to adapt or develop something new