Genome Spot

I recently received access to a new workstation built around the AMD Ryzen Threadripper 2990WX 32-Core Processor. The main use of the system will be to process genome sequencing data. It has 64 threads and the clock speed is 3.2GHz. It should get those data processing jobs done a lot quicker than my other system. Nevertheless I wanted to run some benchmarks to see how much faster it is as well as give it a stress test to see how well it can cope with high loads. This info could also be useful for comparisons in case degradation of the system in the future. Here are the specs of the 3 systems being benchmarked: AMD16: AMD® Ryzen threadripper 1900x 8-core processor - 16 threads@3.8GHz  - 2.4GHz RAM Intel32: Intel® Xeon® CPU E5-2660 16-core processor - 32 threads@3.0GHz - 1.3GHz RAM  AMD64: AMD Ryzen Threadripper 2990WX 32-Core Processor - 64 threads@3.2GHZ - 2.9GHz RAM The command being executed is a pbzip2 of a 4 GB file containing random data using some benchmarking scripts  

The DEE2 project makes freely available bulk data dumps of expression for each of the nine species.  The data is organised as tab separated values (tsv) in "long" format. This is different to a standard gene expression matrix - see below. The long format is prefered for really huge datasets because it can be readily indexed and converted to a database, as compared to wide matrix format.  You'll notice that there are 4 files for each species, with each having a different suffix. They are all compressed with bz2. You can use pbzip2 to de-compress these quickly. *accessions.tsv.bz2 This is a list of runs included in DEE2, along with other SRA/GEO accession numbers.  *se.tsv.bz2 These are the STAR based gene expression counts  in 'long format' tables with the columns: 'dataset', 'gene', 'count'. *ke.tsv.bz2 These are the Kallisto estimated transcript expression counts also in long format *qc.tsv.bz2 These are

Sometimes we need to extract data from an Excel spreadsheet for analysis. Here is one approach using the ssconvert tool. If this isnt installed on your linux machine then you most likely can get it from the package repository. $ sudo apt install ssconvert Then if you want to extract a spreadsheet file into a tsv it can be done like this: $ ssconvert -S --export-type Gnumeric_stf:stf_assistant -O 'separator="'$'\t''"' SomeData.xlsx SomeData.xlsx.tsv You will notice that all the sheets are output to separate tsv files. This approach is nice as it can accommodate high throughput screening, as I implemented in my Gene Name Errors paper a while back. Here is an example of obtaining some data from GEO. $ #first download $ curl 'https://www.ncbi.nlm.nih.gov/geo/download/?acc=GSE80251&format=file&file=GSE80251%5Fprocessed%5FRNA%5Fexpression%5Fmnfyap%2Exlsx' > GSE80251.xlsx $ #now extract $ ssconvert -S --export-type

Dee2.io is a portal for accessing gene expression data derived from public RNA-seq datasets. So far there are over 400k available datasets and its growing every day. While there are existing databases of such as Expression Atlas , Recount2  and ARCHS4 , dee2.io offers a number of unique benefits. For instance, dee2 includes gene-wise counts fron STAR as well as transcript-wise quantifications from Kallisto. There are a few ways you can access these data. Firstly, there is a nice web interface that is mobile friendly. Secondly, there are data dumps available if you are running a large scale analysis.  But the purpose of this post is to demonstrate the improved R interface in action together with SRAdbv2 and statistics with edgeR and DESeq. The official documentation is available on  GitHub . Getting started This tutorial provides a walkthrough for how to work with dee2 expression data, starting with dataset searches, obtaining the data from dee2.io and then performing a differentia

A few updates for DEE2 i would like to share.  I switched over to NameCheap domain name service which appears to be working much nicer than the previous one (HostPapa). The domain name sever change broke the docker image so it was slightly modified and rebuilt.  I've integrated with SRAdbV2, an now there are many more datasets in the queue. I think many of these are small ones related to single cell RNA-seq. I am using as many computers as possible to clear up the backlog. I've noticed a lot of SRA project with one or a few datasets missing, so I have have written a script to identify these and queue them with priority.  The R interface hs undergone several improvements and should be more robust now. A whole bunch of new documentation has been added, including a complete walkthrough starting with SRAdbV2 query, fetching DEE2 data, and differential analysis with edgeR and DESeq. Also bulk data dumps are again available via http. Dat turned out to be too slow and unrelia

For first-pass pathway analysis I find Reactome to be the most useful database of gene sets for biologists to understand. For a long time I have been using Reactome gene sets as deposited to the GSEA/MSigDB website. Recently a colleague pointed me to the gene matrix file offered directly on the Reactome webpage (Thanks Dr Okabe). The latest Reactome gene set matrix file (gmt) can be found at this link  https://reactome.org/download/current/ReactomePathways.gmt.zip   There are some differences. Firstly there are more gene sets in the one from the Reactome webpage (accessed 2018-05-09) $ wc -l *gmt     674 c2.cp.reactome.v6.1.symbols.gmt    2022 ReactomePathways.gmt    2696 total Secondly, there are more genes included in one or more gene sets: $ cut -f3- c2.cp.reactome.v6.1.symbols.gmt | tr '\t' '\n' | sort -u | wc -l 6025 $ cut -f3- ReactomePathways.gmt | tr '\t' '\n' | sort -u | wc -l 10852 And overall there are

As I mentioned in an earlier post , Dat is a new data sharing tool that uses concepts of bittorrent and git to enable peer-to-peer sharing of versioned data. This is cool for sharing datasets that change over time, because when you sync the dataset, only the changes are retrieved, sort of like git. As it uses peer-to-peer technology, it is fairly resilient to node failures as the datasets are mirrored between peers. The "dat publish" command registers the repository on datbase.org, meaning that the files can be retrieved by anyone via a normal browser. To demonstrate, I have released the bulk data dumps from my RNA-seq data processing project, DEE2 , which consists of 158 GB of gene expression data. These data are freely available via a browser at https://datbase.org/dee2/bulk  or by using the dat command-line tool. If you're after a single file, then you can use the following syntax to retrieve over https: wget https://datbase.org/download/<long dat address&g

We know RNA-seq has a number of advantages over array based analyses, but is RNA-seq taking over in terms of number of datasets published? I got curious and thought I'd investigate with some PubMed searches. I searched for "RNA-seq" and "microarray" and downloaded the CSV file which summarises the number of citations per year. As a type of control, I also searched "gene expression". I divided the yearly "RNA-seq" and "Microarray" citation counts by the "Gene expression" counts then multiplied by 1000 to give the numbers seen below. You can see that microarray is still more frequent in PubMed as compared to RNA-seq, but the gap is getting much narrower and the cross will likely occur in the next two years. Next, I will look at the rate of GEO data deposition. (Updates soon)

If you work in genomics, you'll know that sharing large data sets is hard. For instance our group has shared data with our collaborators a number of ways: DVDs, hard drives and flash drives FTP Hightail Google Drive links Amazon links SCP/PSCP rsync But none of these are are ideal as we know data sets change over time and none of the above methods are suited to updating a file tree with changes. If changes occur, then it quickly becomes a mess of files that are either redundant or missing entirely. Copied files could become corrupted. What we need is a type of version control for data sets. That's the goal of dat . So now I'll take you through a simple example of sharing a data set using dat. #Install instructions for Ubuntu 16.04 $ sudo npm cache clean -f $ sudo npm install -g n $ sudo n stable $ sudo npm install -g dat # Files I'm sharing on PC 1: DGE table and 3 genelists (3.4 MB) $ tree . ├── Aza_DESeq_wCounts.tsv └── list     ├── Aza_D

Today I'd like to share some updates on the DEE2 project, which I wrote about  in an earlier post . The  project code can be viewed on github here . Pipeline and images As the pipeline was recently finalised, I was able to roll out the working docker image. To facilitate users without root access, this image was ported to singularity. This took a lot of effort and some expertise from our local HPC team to get things working (many thanks to the Massive/M3 team). The singularity image is available from the webserver ( link ) and instructions for running it are available on github here . I have started testing a heavyweight singularity image, which includes the genome indexes, which will be more efficient for running jobs with large genomes and will make it available once testing is complete. Queue management It may sound simple to write a script to determine which datasets have been completed and add new datasets to the queue but when taking about tens of thousands of datas