-
Notifications
You must be signed in to change notification settings - Fork 24
BLAST
BLAST is one of the most common bioinformatic tools out there, and many biology students learn to run BLAST for individual or small sets of sequences via a web interface: you can find a quick tutorial of web BLAST here. The most popular is probably NCBI BLAST. However, you can perform many more BLAST computations on an HPC.
If you haven't already, log into the ACF, request an interactive session with a computational node, navigate to your folder in the project directory.
Make a directory to hold our blast lesson. Make sure you are in your own folder in our shared project folder.
mkdir blast_examples
cd blast_examples
wget doesn't work on compute nodes. So anytime you need to download data, you will have to do it on the login node. Lets start by downloading some data.
Grab some cow and human RefSeq proteins:
wget ftp://ftp.ncbi.nih.gov/refseq/B_taurus/mRNA_Prot/cow.1.protein.faa.gz
wget ftp://ftp.ncbi.nih.gov/refseq/H_sapiens/mRNA_Prot/human.1.protein.faa.gz
This is only part of the human and cow protein files. check out how big they are:
ls -lh
Now we can get an interactive session on a compute node.
qsub -I -A ACF-UTK0138 -q debug -l walltime=1:00:00,nodes=1:ppn=1,naccesspolicy=singleaccount
Change back to the directory with your blast files
cd /lustre/haven/proj/UTK0138/yourusername/blast_examples
BLAST is available through the module system. This is a way for a computer system to manage many software packages and make them available to users. Lets see what software is available on the ACF:
module avail
We know we want blast, so we can ask if it is available directly
module avail blast
Unfortunately its not the most current version, but it'll do. To use this software, we need to "load" it. This sets up your compute environment to use the software and means you can run it without specifying the path.
module load blast
One easy way to see if a program is working is to look for the version or the help pages.
blastn -version
blastp -help
BLAST has a number of possible programs to run depending on whether you have nucleotide or protein sequences:
- nucleotide query and nucleotide db - blastn
- nucleotide query and nucleotide db - tblastx (includes six frame translation of query and db sequences)
- nucleotide query and protein db - blastx (includes six frame translation of query sequences)
- protein query and nucleotide db - tblastn (includes six frame translation of db sequences)
- protein query and protein db - blastp
The database files are both gzipped, so lets unzip them
gunzip *gz
ls
Take a look at the head of each file:
head cow.1.protein.faa
head human.1.protein.faa
Note that the files are in fasta format, even though they end if ".faa" instead of the usual ".fasta". This NCBI's way of denoting that this is a fasta file with amino acids instead of nucleotides.
How many sequences are in each one?
grep -c '^>' cow.1.protein.faa
grep -c '^>' human.1.protein.faa
Grep? Grep is short hand for "globally search a regular expression and print". So what exactly is a regular expression? Well, its a formal language to define search patterns. In this case, we are asking to search for lines that begin with (the ^ symbol) a greater than sign. Regular expressions (also known as regex) are extremely powerful, and if you are going to be doing a lot of file manipulation and data science, worth learning more about!
This grep command uses the c flag, which reports a count of lines with match to the pattern. In this case, the pattern is a regular expression, meaning match only lines that begin with a >.
This is a bit too big, let's take a smaller set for practice. Let's take the first two sequences of the cow proteins, which we can see are on the first 6 lines
head -n 16 cow.1.protein.faa > cow.small.faa
Now we can blast these two cow sequences against the set of human sequences. First, we need to tell blast about our database. BLAST needs to do some pre-work on the database file prior to searching. This helps to make the software work a lot faster. Use the makeblastdb command:
makeblastdb -in human.1.protein.faa -dbtype prot
ls
Note that this makes a lot of extra files, with the same name as the database plus new extensions (.pin, .psq, etc). To make blast work, these files, called index files, must be in the same directory as the fasta file.
Now we can run the blast job. We will use blastp, which is appropriate for protein to protein comparisons.
blastp -query cow.small.faa -db human.1.protein.faa -out cow_vs_human_blast_results.txt
Take a look at the results in nano. Note that there can be more than one match between the query and the same subject. These are referred to as high-scoring segment pairs (HSPs).
Lets also take a look at the help pages. Unfortunately, there are no man pages (those are usually reserved for shell commands, but some software authors will provide them as well), but there is a text help output
blastp -help
To scroll through slowly
blastp -help | less
To quit the less screen, press the q key.
Parameters of interest: -evalue
( Default is 10?!?) and -outfmt
Lets get only very meaningful matches with a different output format:
blastp \
-query cow.small.faa \
-db human.1.protein.faa \
-out cow_vs_human_blast_results.tab \
-evalue .1 \
-outfmt 7
Whoa!? What are those slashes? In a practical sense, these tell the shell "I'm not done with this command yet". In a literal sense, the forward slash is an "escape" character, and its telling the shell to interpret the newline as just a newline, not a "submit" of the command.
Check out the results with nano.
Tab delimited has these default columns:
qseqid Query sequence ID
sseqid Subject (ie DB) sequence ID
pident Percent Identity across the alignment
length Alignment length
mismatch # of mismatches
gapopen Number of gap openings
qstart Start of alignment in query
qend End of alignment in query
sstart Start of alignment in subject
send End of alignment in subject
evalue E-value
bitscore Bit score
I find it very useful to add in the subject sequence description to the tabular output. If you go through help some more, you will find that you can decide which types of output you would like in the tab delimited file and what order they should be in. For example:
blastp \
-query cow.small.faa \
-db human.1.protein.faa \
-out cow_vs_human_blast_results.tab \
-evalue .1 \
-outfmt "7 std stitle"
Lets try a medium sized data set next
head -n 500 cow.1.protein.faa > cow.medium.faa
What size is this db?
grep -c '^>' cow.medium.faa
Lets run the blast again, but this time lets not have the comment lines, and lets return only the best hit for each query.
blastp \
-query cow.medium.faa \
-db human.1.protein.faa \
-out cow_vs_human_blast_results.tab \
-evalue .1 \
-outfmt "6 std stitle" \
-max_target_seqs 1
I picked "max target seqs" for a reason! It's quite a misleading flag - it finds the number of target sequences but they may not be the best matches. See this blog post for more details..
This is still very inefficient compared to what we can do. Lets imagine needing to BLAST 50,000 sequences against a database of 500,000 proteins (a surprisingly common task!). By the end of the semester, our goal is to be able to:
- Write a short python script to split the 50,000 input sequences into a set of 100 files (each containing 500 sequences)
- Write an ACF submission script that will submit 100 jobs (one for each input file) to the cluster to run in parallel
- Get the results back in a few short hours!