pydoc
google 'python3 list comprehensions`
https://docs.python.org/3/ -> Quick search
Help is available inside python interactive shell
>>>help()Just like for man, help text appears inside a pager like more or less. You can use the following commands:
space -> next page
b -> back a page
return -> next line
/ -> search for a string
q quits the pager
>>> help(str)
Help on class str in module builtins:
class str(object)
| str(object='') -> str
| str(bytes_or_buffer[, encoding[, errors]]) -> str
|
| Create a new string object from the given object. If encoding or
| errors is specified, then the object must expose a data buffer
| that will be decoded using the given encoding and error handler.
| Otherwise, returns the result of object.__str__() (if defined)
| or repr(object).
| encoding defaults to sys.getdefaultencoding().
| errors defaults to 'strict'.
...
| count(...)
| S.count(sub[, start[, end]]) -> int
|
| Return the number of non-overlapping occurrences of substring sub in
| string S[start:end]. Optional arguments start and end are
| interpreted as in slice notation.
...
and dir()
>>> help(str.split)
Help on method_descriptor:
split(...)
S.split(sep=None, maxsplit=-1) -> list of strings
Return a list of the words in S, using sep as the
delimiter string. If maxsplit is given, at most maxsplit
splits are done. If sep is not specified or is None, any
whitespace string is a separator and empty strings are
removed from the result.
(END)
awk is a simple unix utility for reformatting text files. An awk script would look like this
BEGIN { print "File\tOwner"} # block executed before main script
{ print $9 "\t" $3} # main script
END { print " - DONE -" } # block executed after main script
You could run it like this awk table.awk. Each column (whitespace-separated) in the input appears in your script as $1, $2, $3 etc. A bit like sys.argv in python.
Let's ignore the BEGIN and END blocks for now.
How could you take a long file listing and print out the owner of each file?
% ls -l
-rw-r--r-- 1 simonp staff 312 Oct 20 11:05 scope_global.py
-rw-r--r-- 1 simonp staff 201 Oct 20 11:03 scope_global.py~
-rw-r--r-- 1 simonp staff 323 Oct 20 10:40 scope_w_function.py
-rw-r--r-- 1 simonp staff 210 Oct 20 10:33 scope_w_function.py~
-rw-r--r-- 1 simonp staff 5 Oct 15 14:15 test.nt.fa
-rw-r--r-- 1 simonp staff 103 Oct 17 19:27 while.py
-rw-r--r-- 1 simonp staff 160 Oct 17 19:27 while_else.py
Here are the column variables explicitly. This is not shell output. Just a picture.
$1 $2 $3 $4 $5 $6 $7 $8 $9
-rw-r--r-- 1 simonp staff 160 Oct 17 19:27 while_else.py
We want to print the file and the owner. Find the variables. The order can be whatever we want. The awk part would look like this
awk '{print $9 "\t" $3 }'
Note space-separated values would be coded like so
awk '{print $9, $3 }'
How do we get the long listing? ls -l
Put these together with our friend pipe | and add a tab between columns
% ls -l | awk '{print $9 "\t" $3}' # no commas!! why??
scope_global.py simonp
scope_global.py~ simonp
scope_w_function.py simonp
scope_w_function.py~ simonp
test.nt.fa simonp
while.py simonp
while_else.py simonp
Printing files modified on or after the 20th October
% ls -l | awk '$7 >= 20 && $6 == "Oct" {print $9 "\t" $3}'
scope_global.py simonp
scope_global.py~ simonp
scope_w_function.py simonp
scope_w_function.py~ simonp
How can you print the number of records in a fastq file? Each record is 4 lines long.
wc -l mouse_R7385F_*fastq | awk '{print $1/4 "\t" $2}'
85534 mouse_R7385F_smpl1.fastq
688103 mouse_R7385F_smpl10.fastq
100668 mouse_R7385F_smpl2.fastq
802370 mouse_R7385F_smpl3.fastq
407677 mouse_R7385F_smpl4.fastq
Another example that counts how many reads align with score 60 or more (great alignment) to the first 5kb of the reference.
samtools view /Users/simonp/src/samtools-1.14/test/mpileup/mpileup.2.bam | awk '$4 <=5000 && $5 >=60 {print $1 "\t" $2 "\t" $5}' | wc -l
Much simpler than python, put the re between slashes
awk '/li/ { print $2 }' mail-list # match on the whole line
awk '{ if ($1 ~ /J/) print }' inventory-shipped # match on a field (column 1)
ls -l | awk ' /smpl[0-9]/ {print $9}' # match a digit 0,1,2,3 etc
Use awk if you want to reorder columns in a file, do simple filtering and calculations etc. Works great in piped command lines.
Here's a way to save typing
alias is a unix comand that goes in your ~/.profile file (bash) or ~/.zshrc (zsh). Make one with vi if you don't have one already.
alias ll='ls -l'
alias lr='ls -ltrh'To get these changes, source ~/.profile or source ~/.zshrc or open a new window in terminal. Now you can type lr instead of ls -ltrh
Your coding day is time spent doing these things:
- thinking: design
- preparation, testing
- writing code
- debugging
- running code
- thinking: analysis
- more writing, thinking
- report results
Where do you spend most of your time? What can you save time on? The more you plan out coding and check your data, the faster you'll get to the important second half of this list.
- thinking: design Lots of time!
- preparation, testing Lots of time!
- writing code Quick now that you've done the first two
- debugging Quick now that you've done the first two
- running code Very quick
- thinking: analysis Spend lots of time on this and later steps
- more writing, thinking
- report results
Assume your data is corrupted, even if it came from a good colleague. This will stress test your code before you start writing.
Check for consistent numbers of columns in your data, files that end halfway through a line are truncated or corrupted. Is a column always numbers or mixed numbers and text? Be precise about numbers. 2000-3000 is not a number. Nor is 5kb. Do some fields have quotes or other unusual characters, accents? Do the values seem reasonable? Are values for gene lengths between 1,000 and 10,000bp for example?
Data consistency, corruption, sanity checks
NGS data generation: illumina, pacbio
formats - see biopython
(un)compression
filesize and md5 checksums
Say you want to automate blast runs. Here's an example of what we'd run in the terminal.
# first you make a blast database from the fasta file of the sequences
#you want to blast against (target or subject sequences)
makeblastdb -in EcoliO157.uniprot.fa -dbtype prot -parse_seqids
# now you can run a protein blast search with your query ilvG
blastp -query ilvG.bacteria.prot.fa -db EcoliO157.uniprot.fa -outfmt 7 -out ilvG.bacteria.prot.fa.blastp.out -evalue 1e-10
Make sure these commands work ok before you start making a pipeline.
Here are the steps we need our pipeline script to perform:
We need to print a usage message
We need to track when and how the script was run
We need to format the blast database if it hasn't been done already.
We need to run blastp
We need to check blastp ran ok (unix return code)
Print summary table of hits
#!/usr/bin/env python3
import subprocess
import sys
import datetime
import os
# help message
# what's a better way to do this?
if len(sys.argv) < 4:
print(f'Usage: {sys.argv[0]} <query protein fasta> <subject protein fasta> <min E-value>')
exit(1)
# get cmd line params
query = sys.argv[1]
db = sys.argv[2]
evalue = float(sys.argv[3]) # immediately convert to appropriate type
if not query.endswith( ('.fa','.fasta') ):
print('Query input file needs to end with .fa or .fasta')
exit(12)
if not db.endswith( ('.fa','.fasta') ):
print('Subject input file needs to end with .fa or .fasta')
exit(12)
# 2019-10-23 13:49:27.232603
now = str(datetime.datetime.now())
# cut down to 2019-10-23 13:49
now = now[0:16]
#log run command and time/date to screen
print('#' , ' '.join(sys.argv))
print('#' , 'was run on', now)
# see if we have a blast database and make one if not
# let's figure out the next line in detail -> see terminal session
if not os.path.exists(f'{db}.phr'):
makeblastdbcmd = f'makeblastdb -in {db} -dbtype prot -parse_seqids'
makeblastdb_run = subprocess.run(makeblastdbcmd, shell=True , stdout = subprocess.PIPE, stderr=subprocess.PIPE)
# Now we need to check the UNIX return code
# always do this!
# 0 = success
# non-zero =failure
if makeblastdb_run.returncode != 0:
print("FAILED!")
exit(5)
#generate blast output filename
blast_out = query + '.blastp.out'
# run the command
blastcmd = f'blastp -query {query} -db {db} -outfmt 7 -out {blast_out} -evalue {evalue}'
# object is returned after run command
blastcmd_run = subprocess.run(blastcmd, shell=True , stdout = subprocess.PIPE, stderr=subprocess.PIPE)
# Now we need to check the UNIX return code
# always do this!
# 0 = success
# non-zero =failure
if blastcmd_run.returncode != 0:
print("FAILED!")
exit(2)
# now parse results,
homologs = {}
with open(blast_out,'r') as blast_results:
for line in blast_results:
line = line.rstrip()
if line.startswith('#'): # skip comment lines
continue
fields = line.split('\t')
query = fields[0]
subject = fields[1]
evalue = float(fields[10]) # this will be a string because
# we read in from a file
# don't forget to convert to float
# collect hits and evalues into dictionary
if query not in homologs:
homologs[query] = [ (subject, evalue) ]
else:
homologs[query].append( (subject,evalue) )
print('Hit summary')
for query in sorted(homologs):
print('Query:',query)
for data in homologs[query]:
subject,evalue = data
print(f'{subject} E-value={evalue}' )Here are some bare-bones guidelines to get you going.
cutadapt
trimgalore
trimmomatic
fastqc multiqc (aggregate results from bioinformatics analyses across many samples into a single report)
GATK, FreeBayes
Maker (eukaryotes),
Prokka/prodigal (prokaryotes)
Interproscan
sqlite3 is the simplest. It stores your data in a single file. Portable and simple. Gets you up and running quickly.
python has a module
import sqlite3We won't talk about DBs more here, but they are useful for larger data projects. They use their own language: SQL = structured query language.
NCBI (sequences, searching)
Landmark (modle organisms)
Above are better than: nr (proteins), nt (nucleotides) Lots of data, uncurated, complete
Sequence Read Archive (SRA) 454, illumina, short reads
Uniprot (sequences, searching) http://www.uniprot.org Curated, smaller, not as inclusive as nr. Helpful for speeding up analysis: UniRef90 (sequences clustered at 90% identity, which is approximately genus level). Much smaller than full database.
PDB Protein Data Bank For protein structures
Genomes Ensembl, JGI (plants, fungi, bacteria/metagenomes), NCBI genome Organism data bases, beware data quality: some are excellent, some not so well resourced.
import cgi
import cgitb # gives helpful error messages
cgitb.enable()
form = cgi.FieldStorage() # get parametersSee also Flask python library
Run your script with the debugger module pdb for python debugger. Not very sophisticated, but very useful. It starts an interactive debugger that's a bit like the python interactive shell, but you are inside your script.
We were doing this
% python3 while.pyNow we add -m pdb so it becomes
% python3 -m pdb while.py
> /Users/simonp/git/pfb2017/scripts/while.py(3)<module>()
-> count = 0
(Pdb) h
Documented commands (type help <topic>):
========================================
EOF c d h list q rv undisplay
a cl debug help ll quit s unt
alias clear disable ignore longlist r source until
args commands display interact n restart step up
b condition down j next return tbreak w
break cont enable jump p retval u whatis
bt continue exit l pp run unalias where
Miscellaneous help topics:
==========================
exec pdb
(Pdb)
q quits
h gets help
It's a good idea to make alias for python3 -m pdb in .profile. How would we do that?
PyCharm An ok IDE. People also like sublime.
Even though python is much slower than C and C++, is your script running too slowly? How can you tell? Two things to think about
- Is debugging painfully slow? Use the smallest test data sets you can to test and debug your script
- Do you have time to get a cup of coffee while your script is running? If you come back to your script and it's still running, and you're bored, look into speeding it up. Look up profilers, parallelization, other peoples' experiences (seqanswers.com, stackoverflow.com)
Once you are a decent programmer, the speed up you'll get (a few milliseconds) from tinkering with your script (several hours) will not be worth it.