New Core

[John.Sawyer]

Hi all,

I'm starting a new core facility and need some feedback.

When dealing with customers, what are your biggest pain points?

I'd like to learn and start thinking about these. I appreciate your help.

-John

[severin]

Running a Core

I will answer your question indirectly as a core's best practices are meant to minimize conflict between a core and its users. I started my core in September of 2011. Here is what I have discovered so far.

1) Initial Consultation is free, usually no more than an hour.
Your users want to know you are qualified to analyze there data, Here is where you can show them you are.

2) Encourage all users and any potential users to come to you for experimental design advise. At least 3 biological replicates, preferably more where appropriate. Paired end over single end data. I can not think of many, if any, examples where single end data is superior to paired end data.

3) Review grants, provide text and letters of support.

4) Points 1,2 and 3 all will grow your customer base and support your core. So these services should be provided free of charge. Point 2 will make downstream data analysis more productive and results more meaningful.

5) Realize that billing, core managment, pipeline generation and value added scripting will take a significant chunk of your time. Will be doing well if 60% of your time is recovered by data analysis for users of your core.

6) Network with other facility managers that have started up cores or are currently running a core.

7) Social media such as twitter and blogging can be an effective means of identifying useful programs.

8) Track all of your work with a Wiki (I use dokuwiki) and allow users to see every command you used to generate analyzed data. In this way, your users can reproduce your methods if they or one of their students is willing to learn.
It also keeps a record for materials and methods write up, meeting summaries and tracking down errors/problems.

Feel free to email me if you would like to talk more about it over the phone.
[email protected] or [email protected]

Good luck!

[John.Sawyer]

NGS customer consultation

Can you describe the consultation process with customers who don’t know exactly what they should order?

Are most core facility customers thinking of the instrument, # of flow cells and lanes /or/ do they think of read numbers and read length they need (or something else)? We have a few platforms: (Life Tech and Illumina). If a customer comes to me asking for 30 million reads and a 100bp read length, if both instruments can get the job done, does it matter to the customer what instrument I use?

[kmcarr]

2) Encourage all users and any potential users to come to you for experimental design advise. At least 3 biological replicates, preferably more where appropriate. Paired end over single end data. I can not think of many, if any, examples where single end data is superior to paired end data.

There are valid uses for single end data. If your experiment is differential gene expression it is read counts that matter, not total bases. In a model or otherwise well annotated organism the additional information gathered from PE 100 vs SR 50 does not significantly enhance counting reads so is not worth the cost or time. Researchers have tight budgets and if you can get them to trade fewer bp per read for more biological replicates you have helped them design a better experiment. Also, a PE 100 run requires 9-12 days on a HiSeq; an SR 50 only 2.5-3.5 days. Our sequencers are backlogged so being able to shorten run times significantly in cases where it will not impact the integrity of the experiment is a good thing.

[severin]

50bp SE

Fair enough, If you have an organism that is very well annotated AND the organism has minimal repetitive DNA or large homologous gene families that you are interested in then yes 50bp PE data would work perfectly well.

I do tend to think about the vast number of polyploidies found in plants, though even in non-model animals with draft genomes, I still recommend at least 50bp PE. In a well annotated and sequenced bacteria, I think you could make a stronger case for 50bp SE data as there are no splice variants. Keep in mind most down stream count data is based on reads that map to a unique location, PE data can increase the amount of reads that map to a uniq location and 50bp PE is not so much omre than the 50bp SE, though as you mention will take twice as long to generate.

It is not that you can't use 50bp SE data it is that the paired end information provides so much for information. 1) improved mapping, 2) ability to look at splicing information 3) ability to identify large insertion or deletions or transposable elements found in the individual you are sequencing compared with the reference you mapped to.

Cost is always a problem and must be weighed with the advantages/disadvantages of the experimental design.

[severin]

faculty

Can you describe the consultation process with customers who don’t know exactly what they should order?

Are most core facility customers thinking of the instrument, # of flow cells and lanes /or/ do they think of read numbers and read length they need (or something else)? We have a few platforms: (Life Tech and Illumina). If a customer comes to me asking for 30 million reads and a 100bp read length, if both instruments can get the job done, does it matter to the customer what instrument I use?

I have found most faculty come with a biological question they would like to answer. Based on the question they have, I recommend the best platform that will meet their needs and then tell them what other information they can get from the same data.

[John.Sawyer]

user knowledge

Does that mean some faculty are not sure about the numbers of reads or coverage they need for their particular biological questions?

I noticed that there are some general standards:

http://encodeproject.org/ENCODE/protocols/dataStandards/ENCODE_RNAseq_Standards_V1.0.pdf

If its the case that a majority don't know the numbers of reads or coverage they need, do you have certain references you give these faculty members? I'm just trying to equip myself for the types of questions I'll be getting. Thanks!

[severin]

Specialization

Most researchers are specialists in their field and study a very specific question, hopefully something they find full filling and fits within the grant they wrote and was funded.

As managers of core facilities/Bioinformaticians, we are specialists in our field and see a wide variety of projects spanning multiple fields. This includes the technology and software, their ability and limitations. The more we understand the better we are able to provide advise for experimental design and have improved data analysis. We also need a little background in the field from which the data comes, microbial, plant, animal biology.

Now this is not a true dichotomy in the sense that it isn't one or the other. There are many who do both and are successful and many new scientists that are getting education in both. However, there are also many scientists that have been in their field for many years and are embracing the new technologies but look to the bioinformatician for the technical side of it in the same way they might ask a statistician for advise.