Home

Awesome

silent_tools

A bunch of shell utilities for dealing the silent files

# make a silent file
silentfrompdbs *.pdb > my.silent

# ask what's in a silent file
silentls my.silent  

# ask how many things are in a silent file
silentls my.silent | wc -l   

# extract all pdbs from a silent file
silentextract my.silent   

# extract the first 10 pdbs from a silent file
silentls my.silent | head -n 10 | silentextractspecific my.silent    

# extract a random 10 pdbs from a silent file
silentls my.silent | shuf | head -n 10 | silentextractspecific my.silent  

# extract a specific pdb from a silent file
silentextractspecific my.silent name_of_pdb_0001

# produce a scorefile from a silent file
silentscorefile my.silent   

# combine silent files
cat 1.silent 2.silent 3.silent > my.silent  

# ensure all pdbs in silent file have unique names
silentls my.silent | silentrename my.silent > uniq.silent  

# remove _0001 from all the names in a silent file
silentls my.silent | sed 's/_0001$//g' | silentrename my.silent > renamed.silent

# make a new silent file with the first 10 pdbs
silentls my.silent | head -n 10 | silentslice my.silent > new.silent  

# get all sequence from a silent file
silentsequence my.silent > my.seq

# split a silent file into groups of 100
silentsplit my.silent 100

Installation

Installation is simple. Simply clone the repository and add it to your path.

cd ~
mkdir software
cd software
git clone https://github.com/bcov77/silent_tools
cd silent_tools
echo "PATH=\$PATH:$(pwd)"

These commands will print a line that looks like PATH=.... Add this to your .bashrc

Additionally, silentfrompdbs and silentextract require Rosetta. If you have already installed PyRosetta in your default python, then you're already good to go. Otherwise you'll need to make some symlinks.

cd ~/software/silent_tools
# Note that on your system, .linuxgccrelease may be different
#  Ensure that the first argument in these commands actually exists
ln -s /my/path/to/rosetta/main/source/bin/score_jd2.linuxgccrelease score_jd2
ln -s /my/path/to/rosetta/main/source/bin/extract_pdbs.linuxgccrelease extract_pdbs

Refresher on Silent Files

There are many types of silent files, but there is only one type that should ever be used. <b>Binary silent files</b> are the only type that should ever be used.

There are precisely 2 reasons to use silent files:

  1. All of your outputs are in one file (instead of 1M pdbs...)

  2. Silent files load 10X faster than pdbs.

Both of these points are huge. You won't clog up filesystems and you won't have to transfer absurd numbers of files between systems. Additionally, the faster load times make a huge difference when you forget to run a filter and have to load all 1M of your outputs just to spend 0.1 second analyzing each one.

To tell Rosetta to output to a silent file, use this:

rosetta_scripts ... -out:file:silent_struct_type binary -out:file:silent out.silent

This will produce one silent file with all your pdbs.

If you produce many silent files, it is almost always easier to combine them into one giant one.

cat runs/*/out.silent > combined.silent

To tell Rosetta to read from a silent file, use this:

rosetta_scripts ... -in:file:silent_struct_type binary -in:file:silent in.silent

If you only want to work with a subset of the designs in a silent file, the easy solution is to do this:

rosetta_scripts ... -in:file:silent_struct_type binary -in:file:silent in.silent -tags tag1 tag2 tag3 tag4
OR
rosetta_scripts ... -in:file:silent_struct_type binary -in:file:silent in.silent -tagfile tags.list

However, a better way is this:

silentslice in.silent tag1 tag2 tag3 tag4 > tmp.silent
rosetta_scripts ... -in:file:silent_struct_type binary -in:file:silent tmp.silent
OR
cat tags.list | silentslice in.silent > tmp.silent
rosetta_scripts ... -in:file:silent_struct_type binary -in:file:silent tmp.silent

This way, Rosetta doesn't need to read the entire silent file. You only give it the bare minimum. silentsplit is useful here too.

But I don't want to extract silent files

Extracting silent files is so easy though!

silentextract my.silent

But I only want to look at one silent pdb!

silentextractspecific my.silent that_one_I_want_to_look_at

What about a list of pdbs?

cat list_of_tags.list | silentextractspecific my.silent

I would argue to you that silent files provide a more natural way to work with pdbs. Consider a workflow where we make a bunch of outputs and then filter by column 10 in the scorefile.

With silent files:

cat runs/*/out.silent > combined.silent
silentscorefile combined.silent
cat combined.sc | awk '{if ($10 > 100) {print $NF}}' | silentextractspecific combined.silent

Without silent files:

find runs/ -name '*.pdb' > pdbs.list
cat runs/*/*.sc > combined.sc
cat combined.sc | awk '{if ($10 > 100) {print $NF ".pdb"}}' | grep -f - pdbs.list > good_paths.list
cat good_paths.list | xargs cp -t .

Both methods work, but I hope you agree that the one with silent files is cleaner. runs/ may be safely deleted in the silent file case leaving you with a single file representing your whole run.