SEQanswers

Go Back   SEQanswers > General



Similar Threads
Thread Thread Starter Forum Replies Last Post
Hello lovely people! LassInTheLab Introductions 0 10-29-2012 06:34 AM
genome (%) coverage - what people usually get? dukevn Bioinformatics 15 08-15-2010 10:51 AM
Hello sequencing people! yichiao Introductions 0 09-08-2009 07:19 AM

Reply
 
Thread Tools
Old 12-15-2015, 01:20 AM   #1
bi_maniac
Member
 
Location: Madrid, Spain

Join Date: Mar 2014
Posts: 30
Cool Laboratory people vs Computer people

Hi people, I do not want to be polemic. I am a newcomer in bioinformatics and just want you to tell me in an open and frank way how are relationships between lab people and computer people in your organizations. I mean:

Bioinformatic activities are usually performed by lab people (biologist, biochemist, doctors) that have a certain computer knowledge, or computer people that provide that service to lab people?

Many bioinformatics tools run in Linux. My experience shows me that non data-process people are much reticent to abandon Windows and learn to use Linux. Have you solved this problem? How?

Waiting for your feed-back. I hope you find interesting this debate. Thanks a lot!
bi_maniac is offline   Reply With Quote
Old 12-15-2015, 01:35 AM   #2
SylvainL
Senior Member
 
Location: Geneva

Join Date: Feb 2012
Posts: 179
Default

In my organization, many biologists were not that happy with pure (bio)informaticians, because they did not understand each others. When I arrived (I have a PhD and 6years post-doc as wet-biologists, in parallel of bioinformatics since my PhD), they were quite happy.
Actually, I first had to finish some projects (considered as done by the previous bioinformaticians) and it's true that some reports were quite difficult to understand, or even worse, some analyses did not take into consideration the biology behind the data!!!

My feeling is that it requires an effort (which should be from both sides) to make successful interactions...
SylvainL is offline   Reply With Quote
Old 12-15-2015, 03:01 AM   #3
bi_maniac
Member
 
Location: Madrid, Spain

Join Date: Mar 2014
Posts: 30
Default

Hi Sylvaint, I assume that in your case bioinformaticians are like computer consultants. Am I wrong?
bi_maniac is offline   Reply With Quote
Old 12-15-2015, 04:05 AM   #4
GenoMax
Senior Member
 
Location: East Coast USA

Join Date: Feb 2008
Posts: 7,131
Default

As HT sequencing becomes more commoditized there will be commercial and freeware tools that will become more user accessible/GUI friendly e.g. look at the history of microarrays and the associated analytical tools. So for people who don't want to let go of windows there will always be options, albeit limited by what is included in the available programs.

Many new users are appreciating the need to learn unix to get the most out of their data and lots of teaching curricula are including this as an option/requirement.

It is important for the informaticians to emphasize the fact that a lot of what we do generates hypotheses. They are not the final answer in many cases. They still need to be validated by real experiments at the bench, which can only be done by the bench scientists.

So neither computer or lab people are going to solve real experimental problems alone. Collaborations are the way to move science forward setting aside subject expertise egos. Which for some will/may be hard to do.
GenoMax is offline   Reply With Quote
Old 12-15-2015, 04:33 AM   #5
SylvainL
Senior Member
 
Location: Geneva

Join Date: Feb 2012
Posts: 179
Default

Not at all, don't misunderstand me bi_maniac, they are not consultant but my point of view is that bioinformaticians should not be only informaticians, they should have some knowledge in biology as well... So many time, I met informaticians who called themselves bioinformaticians because they were working on biological data, without even understanding the limits of such data...

Last edited by SylvainL; 12-15-2015 at 04:52 AM.
SylvainL is offline   Reply With Quote
Old 12-15-2015, 10:08 AM   #6
Jouni S
Junior Member
 
Location: Cambridge, UK

Join Date: Dec 2015
Posts: 1
Default

Quote:
Originally Posted by SylvainL View Post
Not at all, don't misunderstand me bi_maniac, they are not consultant but my point of view is that bioinformaticians should not be only informaticians, they should have some knowledge in biology as well... So many time, I met informaticians who called themselves bioinformaticians because they were working on biological data, without even understanding the limits of such data...
In different communities, "bioinformatics" may mean quite different things. The most common meanings I've seen are roughly the following:
  1. The study of computational problems inspired by or relevant to the other meanings of "bioinformatics".
  2. The development of computational methods and tools for solving biologically relevant problems.
  3. The study of biology using computational methods.

Most bioinformaticians I know work on one or two of these. It's rare to find a person who's proficient in all three, because most people lack the necessary computational and/or biological background. As a result, there are often huge gaps between the state of the art of the methods used in 1 and 2, and well as in 2 and 3.
Jouni S is offline   Reply With Quote
Old 12-15-2015, 11:40 AM   #7
bi_maniac
Member
 
Location: Madrid, Spain

Join Date: Mar 2014
Posts: 30
Default

Are command line programs, like samtools that difficult to use?

Should lab people study the principles of linux shell?

Should computer people study the principles of biology? To be honest, I have to say that I do not think that this would not be enough.

I think that lab people need only to 'use' programs (which in linux shell is a kind of programming), they do not need to make programs.

I know it is not an easy question. Here the interest of this debate. IMO (in my opinion).

Last edited by bi_maniac; 12-15-2015 at 11:48 AM.
bi_maniac is offline   Reply With Quote
Old 12-15-2015, 05:14 PM   #8
GenoMax
Senior Member
 
Location: East Coast USA

Join Date: Feb 2008
Posts: 7,131
Default

Quote:
Originally Posted by bi_maniac View Post
Are command line programs, like samtools that difficult to use?
No. They are not difficult to use but since the syntax needed to correctly run them is different for each program it will require some investment in time to before reasonably proficient.

Quote:
Should lab people study the principles of linux shell?
Yes. It is not that difficult. This excellent resource from Korf lab is a good place to start but following an instructor is sometimes better: http://korflab.ucdavis.edu/Unix_and_...ent.html#part1

Quote:
Should computer people study the principles of biology? To be honest, I have to say that I do not think that this would not be enough.
I have often found that it is easy for computer science folks to pick up basics of biology needed (than the other way around) for a particular problem. This is where the collaboration becomes important. The experimental colleagues need to be willing to put some time in to bring computer scientists up to speed on the basic biology of experiment at hand.

Quote:
I think that lab people need only to 'use' programs (which in linux shell is a kind of programming), they do not need to make programs.

I know it is not an easy question. Here the interest of this debate. IMO (in my opinion).
Acquiring "applied" bioinformatics expertise (being able to correctly run available analytical tools) is important and lab scientists can certainly acquire the expertise for doing analysis that is straight forward/repetitive. That can allow computer people to focus on more interesting topics.
GenoMax is offline   Reply With Quote
Old 12-15-2015, 05:32 PM   #9
PabloMarin-Garcia
Junior Member
 
Location: Cambridge(uk)

Join Date: Jun 2010
Posts: 7
Default

In my large experience as bioinformatician (starting as a biologist with informatics and statistical skills) I would say that there is a big conceptual gap between lab people and bioinformatician.

I believe that the best approach is to train lab people properly to understand basic bioinformatics tasks and reproducible research and prepare them to collaborate with bioinformatics groups from the neighbourhood. In every lab I have been working at, took me between two and three month to mentor a biologist to create proper tidy data in excel, and teach them how gather and store reproducible data electronically from their experiments that would allow proper bioinformatics analysis (usually this mean automatic pipeline processing not for only a one off statistical analysis). People in the lab think that the data that they provide is ready for analysis but it is not. Usually they don't know about basic things like that the genome have different build, or don't tell you which transcript they use for annotating their variants or forget to annotate the units of measure of the biological variables when you need to aggregate them from different experiments, or have mixed decimal character (spanish/english) or the date column is mixed US/GB english or is in wide format instead of long so you can not group by factors or is not in a 'dataframe format' but several 'table like' pieces of data scattered in a single excel sheet so you need to clean it up first, etc.. Basically 80% of bioinformatics time is spent in 'debugging the data' and prepare it for the analysis. And many times is simple bad science that someone think that with a sound statistical test or fancy graph they can publish. If you ask too many questions the answer would be, 'don't worry my ph student told me that program X can do the analysis'. They forget that programs 'don't DO the analysis', someone DO the analysis WITH a program. Is like when someone 'adjust to the model' meaning adjust the data to the model instead the model to the data :-(.

I would say that the situation is similar to the 80s gap between biologist and chemist when molecular biology was done by 'biochemists' even before this discipline was formally taught. To do molecular biology yo needed to know a lot about chemicals, for example you needed to know the difference between a monovalent or divalent salt or the difference between isopropanol or ethanol for DNA precipitation or when to use TBE or PBE buffer for electrophoresis or how to use a pH meter and how to calibrate in order to prepare your buffers. You needed to understand the basics of chemistry in order to know what you were doing when following a protocol and preparing your reagents for it. Nowadays with all the ready to use kits you can do molecular biology without any knowledge of chemistry (not necessarily a bad thing but it is sad that many people don't know what they are doing or why nowadays).

When I started in molecular biology I read the molbiol bible, the three Maniatis volumes https://books.google.es/books?id=G5RqAAAAMAAJ and this gave me a proper understanding of what I was doing.

In bioinformatics today the situation is very similar. You need to program a pipeline, you need to process text files and dataframes so you need to understand how to store the information in a consistent way for processing it later. You need to understand how data is processed, sorted and filtered and how to do exploratory data analysis (plotting the data and running basic statistic tests and metrics). In the word lab-scientist, the important part of the word is not LAB but SCIENTIST. In todays lab research you don't need to read the Maniatis because the lab kits do it all this for you but instead you need to learn a bit of linux tools for raw text processing and learn to store (or export) your data in .csv and .txt instead of .xls and .doc AND read some good book about bioinformatic or do some of the coursera or edx courses for data-science or bioinformatics. At the end of the day, the lab people that performed the experiment are the ones that better understand the data. All the cleaning and tidying of the data should be done by them as much as possible. The bioinformatician should teach and guide them (or cooperate) until the data is ready for analysis and can be handled safely to the bioinformatics pipeline for analysis. We need to keep in mind that the data would be given back to the lab people as a table or plots that they probably don't understand. Working with them together preparing the data would help us to know how to communicate with them when we need to explain the results tables. And many times even the whole bioinformatics process should be done together with the lab person. A bioinformatician knows about processing the data but the person that designed the experiment know about how to interpret the results and which graphs or filters are the more meaningful for the current study.

In a lab-bioinformatics relationships there are a lot of possible problems to deal with, my decalogue could be:

1- many lab people don't understand how to do a proper experiment design (replicates why for? random allocation?) and the ones that know how to do it don't have the resources for doing it right.

2- Lab people only use excel. Using excel is not a bad thing. Using excel because you don't have any training about how to create 'tidy data' and how to be consistent in nomenclature etc is an atrocious sin (excel is only helping uneducated people to keep going without doing any effort in they education).

3- Informaticians forced to play as bioinformaticians have a hard time understanding the changing nature of biology data and how noisy it is.

4- Some lab people think that bioinformatics is just pressing a big red button and obtaining a pvalor to select below 0.05 (torturing the data or changing software for accomplish this quimera). I have heard many times "I don't need a bioinformatician, I can do the analysis, it is very easy, there is a program for analysing this data". But have you read the manual and do you know how the input data should be prepared, and which are the sensible options of the program given your experiment and the nature of your data? probably not.

5- a biologist need to realize that bioinformaticians are not data magicians. We don't analyze data, we analyze experiments. So the biologist need to be prepared to be asked about experiment design, source of variations, how disperse are the data of the variables they are measuring, which are the technical replicates and the biological ones (usually you need to explain the difference between them, meaning that the experiment design would be flawed by sure) etc.

6- Any one is called a bioinformatician nowadays. You just have to have run some bioconductor scripts or done a basic alignment workflow (without looking really to the sam files or knowing that you can call a complex variant in so many ways ...). I can tell you that bioinformatics is not easy stuff. There is a lot of tricky situations and you need a lot of 'hours of flight' to do a decent job.

7- Bioinformatics, data-anlaysis, biostatistics, computer science... are different things. A scientist depending of their background, can master several of these disciplines but this is something that comes with the person not with the role. A bioinformatician can or can not know about hardware, programming or statistics etc.

8- In the lab the young predoctoral students are assigned to a postdoc to learn the trade. But then a bioinformatician is hired and leave alone. They want a bioinformatician as a service provider or technician but the lab people don't realize that 90% of the things that are going to ask are not pret-a-porter solutions but tailored ones, therefore it is not a service nor a technician task but something to be done by a scientist fellow. This creates the the lonely bioinformatician syndrome https://biomickwatson.wordpress.com/...informatician/ and https://biomickwatson.wordpress.com/...clinical-labs/

9- Bioinformatics analysis are as important as lab work. Sometimes the lab people are acting only as the technicians (prepare the library and run sequencer) but they want to list the variants in a paper and have first author when indeed all the work was done in the bioinformatics realm. https://biomickwatson.wordpress.com/...atics-anymore/

10- Lab people (well, PIs and head of institutes) need to realize that bioinformatics have a huge cost and is not free. http://massgenomics.org/2015/10/ngs-...-not-free.html
PabloMarin-Garcia is offline   Reply With Quote
Old 12-16-2015, 12:20 AM   #10
SylvainL
Senior Member
 
Location: Geneva

Join Date: Feb 2012
Posts: 179
Default

I am fully in agreement with PabloMarin-Garcia.

Genomax, I am not sure whether it is easier to learn biology for a computer person than learn basics of programming for biologists... would say it (again) depends on the person, about its abilities on maths for example. When I joined my institution, so many biologists had absolutely no idea how was perform their analysis. A huge black-box... and worst is that some of them were using a pipeline without even understanding the different options used at each step... In the contrary, I met many bioinformaticians who did not understand the differences (meaning using the same pipeline) for bacteria and human...
SylvainL is offline   Reply With Quote
Old 12-16-2015, 02:12 AM   #11
GenoMax
Senior Member
 
Location: East Coast USA

Join Date: Feb 2008
Posts: 7,131
Default

Quote:
Originally Posted by SylvainL View Post
Genomax, I am not sure whether it is easier to learn biology for a computer person than learn basics of programming for biologists... would say it (again) depends on the person, about its abilities on maths for example.
No argument there. I was going on my personal experience of more years than I care to admit

Before we go too deep into this thread perhaps the broad classification (if that is right word) used by @bi_maniac should be set aside. No one is permanently in one class or the other. We have all gone back and forth (perhaps not strictly in the sense of picking up a pipette and doing a real experiment but thinking about how an experiment can/should be done). Trying to put people into one or the other class is artificially creating this argument.

Quote:
I met many bioinformaticians who did not understand the differences (meaning using the same pipeline) for bacteria and human...
That is unpardonable. They were clearly not "collaborating" with the experimentalists in this case.
GenoMax is offline   Reply With Quote
Old 12-16-2015, 02:28 AM   #12
SylvainL
Senior Member
 
Location: Geneva

Join Date: Feb 2012
Posts: 179
Default

Quote:
Originally Posted by GenoMax View Post
That is unpardonable. They were clearly not "collaborating" with the experimentalists in this case.
I agree but the main problem I identified since I am working full-time as Bioinformatician is the (shared) incapacity of Biologists and Informaticians to speak a common language... resulting in sterile collaborations (not always thankfully). Would always remember one of the Pi going to discuss with a Bioinformatician about one of his analysis concerning circadian-regulated genes. The Bioinformatician proposed to do a Fourier tranform... The Biologist agreed without knowing what it meant. When I explained him the goal was to use a sinus fit in order to find the main pattern, he was quite happy
SylvainL is offline   Reply With Quote
Old 12-17-2015, 12:27 AM   #13
Brian Bushnell
Super Moderator
 
Location: Walnut Creek, CA

Join Date: Jan 2014
Posts: 2,707
Default

Quote:
Originally Posted by PabloMarin-Garcia View Post
(excel is only helping uneducated people to keep going without doing any effort in they education).
Now, now I like Excel!

Otherwise, I pretty much agree with what you said. But, I don't want to put words into anyone's mouth, so I'll state this independently.

In my experience, biologists value other biologists. Biologists in charge do not want to waste their limited grants by hiring computer people when they could hire more biologists. Senior biologists (in my experience) treat bioinformaticians - or, generally, anyone who uses a computer - as technicians who are easily replaced; essential, but disposable.

I think this will change over time as the biologists who currently run labs, but made a name for themselves in pre-computer days, start to retire. That said, I would never recommend a programmer venture into bioinformatics unless they have a strong reason.
Brian Bushnell is offline   Reply With Quote
Old 12-17-2015, 12:39 AM   #14
SylvainL
Senior Member
 
Location: Geneva

Join Date: Feb 2012
Posts: 179
Default

Quote:
Originally Posted by Brian Bushnell View Post
In my experience, biologists value other biologists. Biologists in charge do not want to waste their limited grants by hiring computer people when they could hire more biologists. Senior biologists (in my experience) treat bioinformaticians - or, generally, anyone who uses a computer - as technicians who are easily replaced; essential, but disposable.
I agree with this. Many PI consider Bioinformaticians as the guy who uses softwares, nothing more!!!

But I also agree it is changing... previous sterile collaborations helped some PI to realize they are researchers, not technicians... On the other hand, I also see more often ads for Bioinformatics position with a required bench experience...
SylvainL is offline   Reply With Quote
Old 12-17-2015, 12:56 AM   #15
SylvainL
Senior Member
 
Location: Geneva

Join Date: Feb 2012
Posts: 179
Default

I actually do not like so much the title: laboratory people are not "vs" computer people, neither the opposite. To my point of view, both have to merge in one speciality... (I am speaking mostly about data analyst here, hard bioinformaticians, which develop programs, may still be mostly informaticians... IMO)
SylvainL is offline   Reply With Quote
Old 12-17-2015, 05:58 AM   #16
lre1234
Senior Member
 
Location: US

Join Date: Aug 2011
Posts: 106
Default

As some one who did a solely molecular biology based PhD and has since transferred into the computational world, I think that this site is somewhat biased towards the bioinformatics people. (this is not necessarily a bad thing, and it has been exceedingly helpful to me over the past few years.) But, one thing I have noticed is that many bioinformaticians do not always understand the finer points of a lab experiment. Not in terms of the concept of using biological or technical replicates, but the physical act of doing cell culture, western blots, etc. I have met some people, who would look at a paper in Cell for example, and see 7 figures of western blots, and just think that al you have to do is buy an antibody and it's done, when in reality there is lots of optimization and other work behind all of that. The same goes for the lab people who think that analyzing a number of sequence datasets is just a click away.

So, no 1 person can do it all, and good communication between all parties is need.

As a side and maybe unrelated note, I think that one problem is that in many if not the majority of papers that come out, the data analysis sections are rather poorly described, and just give a reference to an older paper, especially when something custom or unique was done. Maybe there should be better standards on how to present the methods on how data was analyzed, which may give a better explanation to the lab people. The same goes the other way, better description of methods for the lab people. Who knows, this may fix the lack of reproducibility in many papers!
lre1234 is offline   Reply With Quote
Old 12-17-2015, 10:03 AM   #17
Jessica_L
Senior Member
 
Location: Washington, D.C. metro area

Join Date: Feb 2010
Posts: 118
Default

Quote:
Originally Posted by lre1234 View Post
...I think that this site is somewhat biased towards the bioinformatics people. (this is not necessarily a bad thing, and it has been exceedingly helpful to me over the past few years.) But, one thing I have noticed is that many bioinformaticians do not always understand the finer points of a lab experiment.
I was just having a variation on this conversation the other day. I said then, as I'll say now, that I think it's been a natural progression over time. Way back when, none of the protocols were nailed down and Illumina was constantly changing reagents and preps to address performance issues. There was a lot of discussion here about best practices or does-your-data-look-better-with-the-new-chemistry, etc. A lot of those forums settled down as performance issues were worked out.

While a lot of the wet lab work was eventually simplified into kits, the way data gets processed hasn't been that straightforward or static-- you might adjust the way data is trimmed based on your run QC metrics, or you might want to adjust a filter in a variant caller if you're interested in novel variant detection for a particular sample. Questions come up around how to do that, or the best practices for doing so. Then there are people like me that have been doing the bench work for years and there's a lot to learn on the analysis end, so we ask questions. Also, Illumina used to have next to zero technical support for bioinformatics/pipeline problems, so we came here for answers; now most analysis has shifted to open tools like tophat/bowtie/cufflinks or BBTools or SAMTools and if we have issues, we still come here for answers.
Jessica_L is offline   Reply With Quote
Reply

Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off




All times are GMT -8. The time now is 12:29 AM.


Powered by vBulletin® Version 3.8.9
Copyright ©2000 - 2021, vBulletin Solutions, Inc.
Single Sign On provided by vBSSO