FASTA files can be very big and unwieldy, especially if lines are at most 80 characters, one can't speed up browsing them by using less with -S to have one sequence every two lines.

How can I extract just the strain names (or sequence names, i.e. the string on the line starting with >) into a list? I'd like to use seqkit since I'm already familiar with it and know some of the flags and idioms it uses.

I googled the question but couldn't find an answer using seqkit - the examples all use awk, grep, sed or similar, which are alright for copy/pasting but not very memorable.


3 Answers 3


The nice thing about seqkit is that the answer is really simple and memorable:

seqkit seq -n sequences.fasta

This will output a text file containing one sequence name per line:

  • $\begingroup$ Thanks for posting back @CorneliusRoemer thats useful. $\endgroup$
    – M__
    Commented May 30, 2022 at 12:02

You can trivially do this with standard *nix tools. Here are a few options:

sed -n 's/^>//p' file.fa 
awk 'sub(/^>/,"")' file.fa
grep '^>' file.fa | tr -d '>'
perl -ne 'print if s/^>//' file.fa 

The advantage of these is that they will work on any POSIX-compliant system and most other *nix flavors, even if they're not POSIX such as Linux, out of the box without needing to install any external tools.

Another thing to take into account is how fast a command is, especially when dealing with large files. I ran a few tests using the solutions presented here on the human hg38 genome assembly which can be downloaded here and seqkit version 2.2.0 which was downloaded from here. The file was stored locally on my laptop's SSD drive, and the laptop itself is:

$ inxi 
CPU: quad core Intel Core i7-6820HQ (-MT MCP-)
speed/min/max: 2896/800/3600 MHz Kernel: 5.17.5-arch1-1 x86_64
Up: 25d 10h 27m Mem: 15035.9/31958.1 MiB (47.0%)
Storage: 476.94 GiB (103.1% used) Procs: 452 Shell: Bash inxi: 3.3.15

To get more accurate results, I used a little function I have which will run the command you give it a specified number of times and then report the average real time across all runs. Here's the code if you want to try it:


    for com
        printf '\nCOMMAND: %s\n' "$com"
    while [[ $c -lt $max ]]
      let c++;
      ( time -p eval "$com" ) 2>&1 | grep -oP 'real.*?\K[\d\.]+'
        done | 
        awk -vm=$max '{k+=$1}END{print (k/m)}';

And the results. These are the average "real" (i.e. wall clock) times taken across 10 runs:

$ timethis 10 \
  "rg '^>' GCA_000001405.29_GRCh38.p14_genomic.fna | tr -d '>' >/dev/null" \
  "grep '^>' GCA_000001405.29_GRCh38.p14_genomic.fna | tr -d '>' >/dev/null" \
  "sed -n 's/^>//p' GCA_000001405.29_GRCh38.p14_genomic.fna >/dev/null" \
  "perl -ne 'print if s/^>//' GCA_000001405.29_GRCh38.p14_genomic.fna >/dev/null" \
  "./seqkit seq -n GCA_000001405.29_GRCh38.p14_genomic.fna >/dev/null" \
  "awk 'sub(/^>/,\"\")' GCA_000001405.29_GRCh38.p14_genomic.fna >/dev/null" \
  "perl -nle 'push @res, $_  while s/^>//; END {print  join \"\n\",@res}' GCA_000001405.29_GRCh38.p14_genomic.fna"

COMMAND: rg '^>' GCA_000001405.29_GRCh38.p14_genomic.fna | tr -d '>' >/dev/null

COMMAND: grep '^>' GCA_000001405.29_GRCh38.p14_genomic.fna | tr -d '>' >/dev/null 

COMMAND: sed -n 's/^>//p' GCA_000001405.29_GRCh38.p14_genomic.fna >/dev/null 

COMMAND: perl -ne 'print if s/^>//' GCA_000001405.29_GRCh38.p14_genomic.fna  >/dev/null 

COMMAND: ./seqkit seq -n GCA_000001405.29_GRCh38.p14_genomic.fna >/dev/null 

COMMAND: awk 'sub(/^>/,"")' GCA_000001405.29_GRCh38.p14_genomic.fna >/dev/null 

COMMAND: perl -nle 'push @res, 3  while s/^>//; END {print  join "\n",@res}' GCA_000001405.29_GRCh38.p14_genomic.fna 

As you can see above, the grep or rg and tr approaches only took around half a second on average, so that is clearly much faster than any of the other choices. However, seqkit is pretty good since it only took around 4 seconds, faster than any of the other ad hoc solutions.

  • 2
    $\begingroup$ @CorneliusRoemer I guess. I have never used it myself and have been doing bioinformatics for more than 20 years. But by "standard" I mean "standard *nix tools", not bioinformatics tools. I edited to clarify. $\endgroup$
    – terdon
    Commented May 31, 2022 at 8:34
  • 1
    $\begingroup$ @CorneliusRoemer there are weakness on reliance on a developers 'kit'. Flexibility, if you wanted to count the number of pangotype A.1 in SARS-CoV-2, my code is easy to modify to />.*A\.1.*/ and introduce $count++ , whereas you would need to search seqkit to see if the developer had foresight. Bugs: are 'kits' completely bug free, how can you be sure? Upgrades: the promised upgrade ain't ever released (very, very common). Distributing your algorithm: this is a headache because you've now got a dependency which you don't have permission to redistribute. $\endgroup$
    – M__
    Commented May 31, 2022 at 14:17
  • 3
    $\begingroup$ @CorneliusRoemer answers providing different approaches are welcome across the Stack Exchange network (note that this site is not Stack Overflow). Especially when there is already a fine answer with using the requested tool. As for what is memorable, that is entirely personal. You are familiar with seqt so that seems easy and memorable for you. I have been using *nix tools for 20 years, so nothing is more memorable than a simple sed s/foo/bar/. But that's fine! That's the beauty of it: we can all provide answers and people can choose whichever approach works well for them. Everybody wins! $\endgroup$
    – terdon
    Commented May 31, 2022 at 14:51
  • 1
    $\begingroup$ @CorneliusRoemer good catch! The test results I show were using a genome file stored on a network drive, which would of course affect the times. I had assumed that wouldn't be relevant since all tools were accessing the same file over the same network. However, I downloaded a different genome file locally and on my local SSD, seqkit is far more competitive. I'll update in a few minutes with the new test results, but do you have any idea why one tool would be affected by the network I/O times more than another? That's very surprising to me. $\endgroup$
    – terdon
    Commented May 31, 2022 at 18:34
  • 1
    $\begingroup$ @CorneliusRoemer Ha! Serves me right for throwing my age around :) I do indeed have 20 years of experience, but I still come from the world of biology, not programming so there will be holes in my knowledge. I have updated the post with the new results, including all data needed to reproduce them, and seqkit now only loses out to grep + tr but is faster than all the others. So much for my years of experience, I have to eat my words now! $\endgroup$
    – terdon
    Commented May 31, 2022 at 18:54

Or a bit more verbose ...

perl -nle 'push @res, $_  while s/^>//; END {print  join "\n",@res}' myfile.fa

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.