Tool:Building a new NGS analysis platform, we would love your feedback (RNA-seq)

Question

[Keith]

Hi all,

Long time lurker, finally breaking the silence to get some feedback from the bioinformatics community. I'm a former post-doc who decided to leave the academic world to help my colleagues work more efficiently with their NGS data.

Introducing, Stirplate.io an automated data analysis tool built for biologists with no experience with bioinformatics. Stirplate allows those with no experience to run their analysis without touching a line of code. It's similar to DNAnexus and Galaxy, but we've tried to make the process even more simple and fast (RNAseq data in under 24 hours for most).

So far, we've built 1 full protocol into the system (differential analysis for RNAseq using DeSeq2) and we would love your feedback before moving forward. We would also love to hear what you, as a community, are interested in seeing in the future. We've all dealt with horrible software in the past, and we really want to make something that makes your job easier.

Why is this different from tools like Galaxy or DNAnexus? We've tried to make the process as frictionless as possible. For example, when you upload data, you can upload any kind of file, in any format (tar.gz, zip, bz2 etc), even if the fastq files are nested within the archive.

We also know that data visualization tends to be a issue for some biologists, so we built interactive graphs that allow you to manipulate data in real time, without having to go back into R and re-plot the data.

The only thing the researcher needs to know is which fastq files match up to their sample names and we take care of the rest (alignment, counting via HTseq, differential analysis, and interactive data visualization.) We also host data for 1 year after you run it, and can continue to host your data as long as you like at a small fee to cover our costs.

We are just getting started and would love your feedback, so we are offering 10 free samples for RNAseq analysis on our system. If you have any issues, you can reach out to me at [e-mail] or post it here so we can discuss as a group.

Thanks!

Keith

Answer 1

[Damian Kao]

Is there a demo of the UI on the site? If not, It would be nice if we can see some kind of demo of the UI. Maybe setup a demo where user can submit some internal mock data and press a few buttons to get pre-processed results.

As a side note, philosophically, I don't completely agree with what you wrote in the "about us" section. I think scientists should try to understand the ins and outs of all their data analysis. I understand that learning the practical steps of how to run something on the command line or changing data formats is probably not that important, but I do hope scientists should at least learn the theory/concept behind the analysis. There are people who perform differential expression analysis without really understanding what the results mean, ultimately leading to wrong conclusions. I hope these services don't perpetuate that.

Answer 2

[Keith]

Hi Damian,

Thanks for the feedback! If you to test the site yourself, I'd be happy to send over some test data to work with, it takes about 10-15 minutes to run through(total). Let me know and I'll post up a link to the file you can test with. Alternatively, feel free to run your own data through. We have indexes for Human, Mouse, Rat, and C. Elegans. We will be making a video sometime soon.

Re: The philosophical side of data analysis, I agree that people should know what is going on with their data. I've seen plenty of bad studies published because people had no idea what they were doing with their analysis. As we grow, we will build in more and more "best practices" into the system to avoid people just putting in garbage and getting garbage data out. For example, we check samples after they go through alignment to make sure they mapped well. If things look bad, we automatically provide a description of what could have happened to cause that problem and stop the analysis. In the future, we will add automatic trimming, best statistical practices and more.

We can't avoid bad research getting out there, but being that everything is controlled with best practices built in, we hope that we can avoid having people publishing bad data due to their poor understanding of the tools currently available.

[Damian Kao]

Something I don't see in most of these web NGS services is an educational component. Probably because that's not really the point of the service and inclusion of educational material would be very time consuming I am sure.

However, It would be cool if during the analysis pipeline, a conceptual description of what is going on is accompanied with each intermediate result. So the user can really understand the analysis procedure and maybe even change parameters accordingly. I wonder if this can be a feature that draws in users.

[Keith]

We actually do have some of that built into the system. We explain wha a PCA plot and a Heat-map is actually showing you as a well as a brief description of the DeSeq Protocol and what it really does. We have intentions of expanding this to each of the protocols that we roll out in the future as well. We also plan on writing up a brief methods section for you, this way writing the paper is much easier.

Answer 3

[Soorni, Aboozar]

Hi

Great, I am a new person in this field and I entirely agree with the philosophical side of data analysis. Before than every things I always think about the aim and choice best method to getting enough and good data. Oftentimes I think about coverage (do I have enough data for analysis or NOT). I don't know exactly but just as an offer, probably you can create a part to estimate coverage and calculate number of reads to having a background about existence enough data base on genome size and others parameters.