[dahlo]

What still amazes me is that people who are prepared to spend tens of thousands of pounds on sequencing won't stump up a thousand pounds for a beefy PC to help with the analysis. It's painful to see a postdoc wasting their time watching analyses run slowly on a 4 year old PC, which could be done in a fraction of the time if they spent a modest amount on a new machine.

I am also amazed by this fact. Even whole genome sequencing of 2-3 samples is cheap compared to having 1 person employed for a year. Still some people reason that time is something that can be wasted without thinking twice..

[simonandrews]

True that may be, but as NGSfan pointed out, the analysis scene is constantly changing. Say that it takes 6-12 months to make a tool user friendly. A new and better tool will come out about as often, so the user-friendly GUIs will always be one generation behind.

Conversely the problem with always using the new and shiny command line tools is that people too often blindly assume that the results they produce must be reliable because they're using the latest and greatest thing.

The benefits of having a nice GUI are that it allows you to visually explore your data so you can put the analysis results you have into some kind of context. You can do this with command line tools, but I guarantee that most people who run them never bother. Personally speaking I'd rather have a simpler analysis which gives me results I can understand and trust over a more complex analysis which gives me some more hits, but where I'm just trusting that the author's assumptions about the data are valid in my case. There are times when getting every last hit you can out of your data is of primary importance, but a lot of the time people only follow up the really strong hits anyway, and being able to understand where those hits came from and trust them is a big advantage.

My time is pretty evenly split between working on a command line and in a GUI, but I'm very confident in saying that my understanding of our data has been advanced far more by exploring it in a GUI than by looking at the
output of command line tools.

[ETHANol]

Novice, I think you think the problem is a little more simple then you think. Imagine giving a 'bioinformatician' a set of pipetmen and asking him to plan an RNA-seq experiment, generate the material, collect the samples and prepare the RNA/libraries. You give him some ready made kits with manuals and tell him to go at it.

Take some time to learn what you need to to process the data correctly. We are not talking about a lot of time, but you will be better off and more efficient if you do. Biology is getting more and more computational, learn and evolve with it, it will make you a better scientist.

Cloud computing, I don't get it. A new PC with a lot of RAM and two big hard drives is so much easier.

[1230Rock]

Bumping to ask a seemingly simplistic question:
I am a biologist with the desire to learn all there is to know about NGS analysis, but my problem has been finding a good source of information that starts at the VERY beginning. If the criticism is biologist should learn to think more like bioinformaticians, then why has it been so difficult for me to do just that? Am I just looking in the wrong places? Is there a web course out there that will teach me the ins and outs of the different alignment programs and analysis softwares using the cloud or otherwise?

[radekchr]

Bumping to ask a seemingly simplistic question:
I am a biologist with the desire to learn all there is to know about NGS analysis, but my problem has been finding a good source of information that starts at the VERY beginning. If the criticism is biologist should learn to think more like bioinformaticians, then why has it been so difficult for me to do just that? Am I just looking in the wrong places? Is there a web course out there that will teach me the ins and outs of the different alignment programs and analysis softwares using the cloud or otherwise?

Hey 1230Rock,

Here at MSU, there is a 2 week summer program for biologists that goes over the basics of a lot of handling of next-gen sequencing data and using Amazon cloud computing.
Here's a link to last year's course, which also provides a link to the exercises we did last year. Not sure this starts at the VERY beginning, but I thought it was useful nonetheless!

http://bioinformatics.msu.edu/ngs-summer-course-2011

[gringer]

These softwares are supposed to be made for people with no prior knowledge of bioinformatics, or at least so does the companies behind it claim, but nope. They talk about the algorithms and explain them in a way so that a biologist like me gets lost half the way (this is especially true about the Avadis). And when you finally have passed the aligning step, it needs a whole education to understand and interpret your results.

For someone to use the high-cost commercial bioinformatics suites effectively (where by high-cost, I mean similar in cost to a bioinformatician's salary), they need very similar skills to what would be required to run low-cost (or FOSS) programs with similar effectiveness.

The programs are not particularly easy to use for complete novices, and not customisable enough that a bioinformatician would be able to set things up so that they are easy to use. Therefore, use of these programs needs substantial amounts of hand-holding by someone with fairly advanced knowledge of bioinformatics (i.e. you would need a double-bioinformatician salary for running these programs).

Given that researchers will typically want to try a few things that haven't been done before, I find it doubtful that a software suite could be designed that will work for almost everyone.

[steven]

Does a pilot have to be a mechanical engineer?

Biology will eventually turn into an information science and if you want to do research you will have to use computers.
Harsh words but this is coming from someone like you who started as a biologist and fought his way into bioinformatics. It's not easy, but for your own career I would strongly urge you to learn - otherwise you will be forever dependent on others to do your analysis - and that means you will be waiting.

I strongly disagree with this.

Biology is an experimental science. The fact that you are a fan of biotechnology and computers -as many of us by the way- does not reduce biology to NGS data analysis. How self centred is this!

Although I also started as a biologist and became a bioinformatician, I do not pretend everyone has to do the same. Either you have infinite learning capacities -good for you then- or you just cannot be an expert in everything. At some point you have to choose. Of course the best is to learn and know about everything, but that is simply not possible in real life. Absolute independence is pure utopia. Unless you work alone in your lab, chances are you will always depend on some collaborator at some point, whether it is to collect samples, set up an experiment, write or review an article, use a machine, take pictures, export the data, get a grant.. why should this become a major issue when it comes to analyzing the data?

Look at it the other way around: as a bioinformatician, I am 100% dependent on biologists to generate some interesting data, and I am fine with that. I won't come back to the bench to try running the experiment myself simply because I am "waiting". First I prefer to have it done by an expert. Second I am never "waiting". To use your words, if a project involves that many people are just "waiting" at some point maybe it shouldn't have been given the funding..

[NGSfan]

I strongly disagree with this.

Biology is an experimental science. The fact that you are a fan of biotechnology and computers -as many of us by the way- does not reduce biology to NGS data analysis. How self centred is this!

I am a biologist by training first. I know and understand everything from beginning (the molecular biology of sample prep) to the end (computational analysis). My graduate school training emphasized that I learn BOTH in order to be competitive and competent.

Analytical chemists and experimental physicists also do experimental science but do you think they just hand off the data from their experiment to another "computational" person to analyze for them? Why is it that biologists think they can get away with just knowing the molecular biology and biochemistry and not understand the computational analysis? (And vice-versa - bioinformaticians that don't have a clue about the biology and treat the data no different than any other data and are completely oblivious to biases in molecular biology sample preps!)

Although I also started as a biologist and became a bioinformatician, I do not pretend everyone has to do the same. Either you have infinite learning capacities -good for you then- or you just cannot be an expert in everything. At some point you have to choose. Of course the best is to learn and know about everything, but that is simply not possible in real life. Absolute independence is pure utopia. Unless you work alone in your lab, chances are you will always depend on some collaborator at some point, whether it is to collect samples, set up an experiment, write or review an article, use a machine, take pictures, export the data, get a grant.. why should this become a major issue when it comes to analyzing the data?

OK maybe I come across as arrogant. It is not about "thumping one's chest" and declaring they are master of the universe. But you should strive to know as much as you can and at the very least - be competent enough to analyze your own data!! It's fine if someone does the BLAST for you or the BWA alignment for you - but the expectation , in my experience , has been that all the analysis and interpretation is to done by the bioinformatican. The biologist see it as "just pushing a button".. yet they take for granted the effort it takes to make sure there are low false positives in those results.

This is a typical problem in biologist - bioinformatician collaborations.

I ask you this - If you are just producing the data and handing it off to someone else to analyze for you, who should get the first authorship on the paper? In my opinion the data producer is just a technician... amirite?

There are plenty of grad students and postdocs out there demanding they get first authorship for spending 6 months to produce the data, but cannot analyze it themselves! Is that not ridiculous? How can you call yourself a scientist if you cannot analyze and interpret the data? or know the quality of your data?

Look at it the other way around: as a bioinformatician, I am 100% dependent on biologists to generate some interesting data, and I am fine with that. I won't come back to the bench to try running the experiment myself simply because I am "waiting". First I prefer to have it done by an expert. Second I am never "waiting". To use your words, if a project involves that many people are just "waiting" at some point maybe it shouldn't have been given the funding..

What if the experiment doesn't work, and you think you can trust the "expert". Often the "expert" is a technician or a grad student... maybe a postdoc doing a new protocol for the first time. How would you know what questions to ask if something goes wrong if you don't know the molecular biology behind the sample preparation and don't have experience doing PCR / ligations / etc ??

I'm sorry but maybe my training was very demanding and in my circles to just throw up your hands and say "well I'm not a biologist / bioinformatician don't ask me" was seen as an excuse that was not acceptable.

[mathew]

After seeing the ad of Avadis NGS on Seqanswer page I tried to use its trial version but they never send an email- either their system is messed up or they just advertise. Then I came to know from couple of my colleagues that they either did not get any sort of trial of this tool (no email) form last 2-3 months. I am not sure then why they advertise, unless they are willing to check their system

Thanks

[Buzz0r]

FYI, Illumina already launched their own cloud solution called BaseSpace. It includes many 3rd party softwares and is opening up completely by the end of the year. Might be worth checking it out!

http://blog.basespace.illumina.com/

Cheers...

[rfilbert]

Partek vs. CLC vs. AVADIS vs. Galaxy

Hi All,

Well I have finished evaluating these 4 options for analysis of NGS data and I decided to go with Partek Flow. The reasons:

1. For Galaxy, you really need to be a certified bioinformaticist to know how to use it - this is not a tool for a biologist like myself.
2. CLC is expensive, and does not have the downstream analysis capabilities that AVADIS or Partek have.
3. AVADIS has too many bugs. I just could not get my analysis done.
4. Partek Flow was the most user friendly software I've ever seen and their sales and technical support team were also far better than the others.

YMMV
Best,
Robert

[mathew]

Galaxy

I will vote for Galaxy as it is very user friendly and provide you the control on your analysis to certain extent not just a black box. If one invest some time one can get into through adjustment of various parameters. I have seen over time that it is always risky to follow blindly a commercial tool.... It is simply too risky.
My thoughts only

[rfilbert]

I have used Galaxy and I have used Partek Flow, and in my opinion Partek Flow is much easier to install and use than Galaxy. It uses both Partek algorithms and open source (e.g. Bowtie/Tophat, BWA, GATK, etc.) but somehow runs them faster than Galaxy. It also has a lot that Galaxy lacks, such as controlled vocabularies, outstanding graphics, etc. For example, I couldn't do the downstream analysis needed for my RNA-Seq study in Galaxy. They also have a dedicated technical support team I can call for help. I'm not sure what you mean by "blindly following a commercial tool"? What is the difference between that and blindly following Galaxy?

[gringer]

Has anyone used Galaxy for NGS analysis from a high-throughput illumina sequencer? I would expect that the browser-based system would choke on the 50-200GiB BAM files that would be produced from a 200-1000M read run.

[rfilbert]

Both Partek Flow and Galaxy are browser based. Being browser based means they both are client/server architectures. It is the server that would need to choke on large data, not the browser. So it depends on the server hardware and how efficiently each system utilizes that hardware. As I mentioned, I have used both for fairly large studies and Flow ran the same aligners significantly faster than Galaxy.