Physical-chemical property predictors as command line tools

The Instant JChem Suite, from ChemAxon, is a fantastic set of software designed for Chemists. It allows easy and simple database management to store both chemical and non-chemical data. It also contains a plethora of physical-chemical prediction and visualisation tools that can be utilised by chemists and computational based scientists alike.

I personally believe that the hidden gem within the suite is the availability of these predictive tools in the command line, found within the ChemAxon Calculator (cxcalc). In addition, calculator plug ins have also been developed by external developers. This allows you to incorporate the powerful predictive tools of ChemAxon into your larger workflows, with a little scripting.

For example, it can be used to predict the dominant protonation state of a ligand before use in MD or docking studies, with the majormicrospecies tool. You can input and output all major file types including SDF, PDB and MOL2 using commands such as:

cxcalc [Input_File].sdf -o [Output_File].sdf majormicrospecies -H [pH] -f [Output_File_Type]:H

You can easily find the calculator plugins available and how to construct input commands using the cxcalc –h command, or the available online information. I would thoroughly recommend looking at the tools available and how you could incorporate them into your workflows.

Counting Threads

When someone talks about “counting threads” the first thing that you think of is probably shopping for bed sheets. But this post is not about the happy feeling of drifting off to sleep on smooth, comfortable Egyptian cotton. This post is about that much less happy feeling when you want to quickly run a bit of code on a couple of data sets to finish the results section of a thesis chapter, and you see this:

Luis blocking the server again....

Luis blocking the server again….

Obviously someone (*cough*Luis*cough*) is having some fun on the server without nice-ing their code to allow people who are much less organized than they should be (*cough*me*cough*) to do a quick last-minute data analysis run.

The solution: confront the culprit with their excessive server usage (Note: alternatives include manual server restart with a power cord to make the world your enemy – not recommended).

So… now we just need to find out how much of the server “ospina” is using. Screenshots won’t convince him… and we can’t take enough screenshots to show the extent of the server-hogging with his 1000s of processes anyway. We need to count…

Luckily there is a handy function to find out information about processes called pgrep. This is basically a ‘ps | grep’ function which has a bunch of options to reflect the many ways it can be used. We see opsina is running R, so here goes:

pgrep -c R

The -c flag counts processes and the pattern matches the command name that was run. But yeah, it turns out this wasn’t the best idea ever. A lot of people are running R (as might be expected in the Statistics Department), and you get a number that is really too high to be likely. We need to be more specific in our query, so let’s go back to the ps command. Second attempt:

ps -Af | grep ospina | wc -l

What we’re doing now is first showing all processes that are run on the server (ps -A) also showing details of the command run and who ran it (-f flag). Then we’re finding the ones that are labelled with our server culprit (grep ospina) and counting the lines we find. There are annoyingly still a few problems with this approach.

  1. We just ran this command on the server and thus will count a command like grep –color=auto ospina,
  2. User “ospina” is probably running a few more things than just his R command (like ssh-ing into the server and maybe a couple of screens)
  3. We get a number than looks far lower than what we expected just by visual inspection.

So… what happened? We can fix problems 1 and 2 by just piping to a further grep command. But problem 3 is different. As it turns out, our culprit is running multiple threads from the same process (which is also why you find so many chrome instances on htop for example). We just counted processes, when really the server is being occupied by his multi-threading exploits. So… if you want to back up your complaint with a nice number, here’s your baby:

ps -ALf | grep opsina | grep R-3.3 | wc -l

The -L flag displays all threads instead of only the processes. I further used R-3.3 as it turns out he is using a specific version of R, which I can use to specify this command. Otherwise it also helps to use inputs arguments to functions to search against. If your fingers get too tired to press the shift-key that often, ps -ALf is equivalent to ps -eLf.

For now: moan away, folks!

 

Disclaimer: Any scenarios alluded to in the above text are fictitious and do not represent the behaviour of the individuals mentioned. Read: obviously I do not do my analysis last-minute.

The Protein World

This week’s issue of Nature has a wonderful “Insight” supplement titled, “The Protein World” (Vol. 537 No. 7620, pp 319-355). It begins with an editorial from Joshua Finkelstein, Alex Eccleston & Sadaf Shadan (Nature, 537: 319, doi:10.1038/537319a), and introduces four reviews, covering:

  • the computational de novo design of proteins that spontaneously fold and assemble into desired shapes (“The coming of age of de novo protein design“, by Po-Ssu Huang, Scott E. Boyken & David Baker, Nature, 537: 320–327, doi:10.1038/nature19946). Baker et al. point out that much of protein engineering until now has involved modifying naturally-occurring proteins, but assert, “it should now be possible to design new functional proteins from the ground up to tackle current challenges in biomedicine and nanotechnology”;
  • the cellular proteome is a dynamic structural and regulatory network that constantly adapts to the needs of the cell—and through genetic alterations, ranging from chromosome imbalance to oncogene activation, can become imbalanced due to changes in speed, fidelity and capacity of protein biogenesis and degradation systems. Understanding these complex systems can help us to develop better ways to treat diseases such as cancer (“Proteome complexity and the forces that drive proteome imbalance“, by J. Wade Harper & Eric J. Bennett, Nature, 537: 328–338, doi:10.1038/nature19947);
  • the new challenger to X-ray crystallography, the workhorse of structural biology: cryo-EM. Cryo-electron microscopy has undergone a renaissance in the last 5 years thanks to new detector technologies, and is starting to give us high-resolution structures and new insights about processes in the cell that are just not possible using other techniques (“Unravelling biological macromolecules with cryo-electron microscopy“, by Rafael Fernandez-Leiro & Sjors H. W. Scheres, Nature, 537: 339–346, doi:10.1038/nature19948); and
  • the growing role of mass spectrometry in unveiling the higher-order structures and composition, function, and control of the networks of proteins collectively known as the proteome. High resolution mass spectrometry is helping to illuminate and elucidate complex biological processes and phenotypes, to “catalogue the components of proteomes and their sites of post-translational modification, to identify networks of interacting proteins and to uncover alterations in the proteome that are associated with diseases” (“Mass-spectrometric exploration of proteome structure and function“, by Ruedi Aebersold & Matthias Mann, Nature, 537: 347–355, doi:10.1038/nature19949).

Baker points out that the majority of de novo designed proteins consist of a single, deep minimum energy state, and that we have a long way to go to mimic the subtleties of naturally-occurring proteins: things like allostery, signalling, and even recessed binding pockets for small moleculecules, functional sites, and hydrophobic binding interfaces present their own challenges. Only by increasing our understanding, developing better models and computational tools, will we be able to accomplish this.

A beginner’s guide to Rosetta

Rosetta is a big software suite, and I mean really big. It includes applications for protein structure prediction, refinement, docking, and design, and specific adaptations of these applications (and others) to a particular case, for example protein-protein docking of membrane proteins to form membrane protein complexes. Some applications are available in one of the hassle-free servers online (e.g. ROSIE, Robetta, rosetta.design), which might work well if you’ve got just a few tests you would like to try using standard parameters and protocols. However, it’s likely that you will want to download and install a version if you’re interested in carrying out a large amount of modelling, or using an unusual combination of steps or scoring function. This is not a trivial task, as the source code is a 2.5G download, then your machine will be busy compiling for some time (around 5 hours on two cores on my old laptop). Alternatively, if the protocols and objects you’re interested in are part of PyRosetta, this is available in a pre-compiled package for most common operating systems and is less than 1G.

This brings me to the different ways to use Rosetta. Most applications come as an executable which you can find in Rosetta/main/source/bin/ after completing the build. There is documentation available on how to use most of these, and on the different flags which can be used to input PDB structures and parameters. Some applications can be run using RosettaScripts, which uses an xml file to define the protocol, including scoring functions, movers and other options. In this case, Rosetta/main/source/bin/rosetta_scripts.* is run, which will read the xml and execute the required protocol.

screenshot-from-2016-09-14-19-19-28

An example RosettaScript, used for the MPrelax protocol

PyRosetta is even more flexible, and relatively easy to use for anyone accustomed to programming in python. There are python bindings for the fast C++ objects and movers so that the increased usability is generally not greatly compromised by slower speeds. One of the really handy things about PyRosetta is the link to PyMOL which can be used to view the trajectory of your protein moving while a simulation is running. Just add the following to your .pymolrc file in your home directory to set up the link every time you open pymol:

run /PATH/TO/PYROSETTA/PyMOLPyRosettaServer.py

When it comes to finding your way around the Rosetta package, there are a few things it is very useful to know to start with. The demos directory contains plenty of useful example scripts and instructions for running your first jobs. In demos/tutorials you will find introductions to the main concepts. The demos/protocol_capture subdirectory is particularly helpful, as most papers which report a new Rosetta protocol will deposit here the scripts required to reproduce their results. These may not currently be the best methods to approach a problem, but if you have found a research article describing some results which would be useful to get for your system, they are a good starting point to learn how to make an application work. Then the world is your oyster as you explore the many possible options and inputs to change and add!

What happens to the Human Immune Repertoire over time?

Last week during the group meeting we talked about a pre-print publication from the Ippolito group in Austin TX (here). The authors were monitoring the antibody repertoire from Bone Marrow Plasma Cells (80% of circulating abs) over a period of 6.5 years. For comparison they have monitored another individual over the period of 2.3 years. In a nutshell, the paper is like Picture 1 — just with antibodies

This is what the paper talks about in a nutshell. How the antibody repertoire looks like taken at different timepoints in an individual's lifetime.

This is what the paper talks about in a nutshell. How the antibody repertoire looks like taken at different timepoints in an individual’s lifetime.

.

The main question that they aimed to answer was: ‘Is the Human Antibody repertoire stable over time‘? It is plausible to think that there should be some ‘ground distribution’ of antibodies that are present over time which act as a default safety net. However we know that the antibody makeup can change radically especially when challenged by antigen. Therefore, it is interesting to ask, does the immune repertoire maintain a fairly stable distribution or not?

Firstly, it is necessary to define what we consider a stable distribution of the human antibody repertoire. The antibodies undergo the VDJ recombination as well as Somatic Hypermutation, meaning that the >10^10 estimated antibodies that a human is capable of producing have a very wide possible variation. In this publication the authors mostly focused on addressing this question by looking at how the usage of possible V, D and J genes and their combinations changes over time.

Seven snapshots of the immune repertoire were taken from the individual monitored over 6.5 years and two from the individual monitored over 2.3 years. Looking at the usage of the V, D and J genes over time, it appears that the proportion in each of the seven time points appears quite stable (Pic 2). Authors claim similar result looking at the combinations.  This would suggest that our antibody repertoire is biased to sample ‘similar’ antibodies over time. These frequencies were compared to the individual who was sampled over the period of 2.3 years and it appears that the differences might not be great between the two.

How the frequencies of V, D  and J genes change (not) over 6.5 years in a single individual

How the frequencies of V, D and J genes change (not) over 6.5 years in a single individual

It is a very interesting study which hints that we (humans) might be sampling the antibodies from a biased distribution — meaning that our bodies might have developed a well-defined safety net which is capable of raising an antibody towards an arbitrary antigen. It is an interesting starting point and to further check this hypothesis, it would be necessary to carry out such a study on multiple individuals (as a minimum to see if there are really no differences between us — which would at the same time hint that the repertoire do not change over time).

 

Processing large files using python: part duex

Last week I wrote a post on some of the methods I use in python to efficiently process very large datasets. You can find that here. Roughly it details how one can break a large file into chunks which then can be passed onto multiple cores to do the number crunching. Below I expand upon this, first creating a parent class which turns a given (large) file into chunks. I construct it in a manner which children classes can be easily created and tailored for specific file types, given some examples. Finally, I give some wrapping functions for use in conjunction with any of the chunkers so that the chunks can be processed using multiple cores.

First, and as an aside, I was asked after the previous post, at what scale these methods should be considered. A rough answer would be when the size of the data becomes comparable to the available RAM. A better answer would be, when the overhead of reading each individual line(/entry) is more than the operation on that entry. Here is an example of this case, though it isn’t really that fair a comparison:


>>> import timeit,os.path
>>> os.path.getsize("Saccharomyces_cerevisiae.R64-1-1.cds.all.fa")
10095955
>>> timeit.timeit('f = open("Saccharomyces_cerevisiae.R64-1-1.cds.all.fa");sum([l.count(">") for l in f])',number=10)
0.8403599262237549
>>> timeit.timeit('f = open("Saccharomyces_cerevisiae.R64-1-1.cds.all.fa");sum([c.count(">") for c in iter(lambda: f.read(1024*1024),"")])',number=10)
0.15671014785766602

For a small 10MB fasta file, we count the number of sequences present in a fifth of the time using chunks. I should be honest though, and state that the speedup is mostly due not having the identify newline characters in the chunking method; but nevertheless, it shows the power one can have using chunks. For a 14GB fasta file, the times for the chunking (1Mb chunks) and non-chunking methods are 55s and 130s respectively.

Getting back on track, let’s turn the chunking method into a parent class from which we can build on:


import os.path

class Chunker(object):

    #Iterator that yields start and end locations of a file chunk of default size 1MB.
    @classmethod
    def chunkify(cls,fname,size=1024*1024):
        fileEnd = os.path.getsize(fname)
        with open(fname,'r') as f:
            chunkEnd = f.tell()
            while True:
                chunkStart = chunkEnd
                f.seek(size,1)
                cls._EOC(f)
                chunkEnd = f.tell()
                yield chunkStart, chunkEnd - chunkStart
                if chunkEnd >= fileEnd:
                    break

    #Move file pointer to end of chunk
    @staticmethod
    def _EOC(f):
        f.readline()

    #read chunk
    @staticmethod
    def read(fname,chunk):
        with open(fname,'r') as f:
            f.seek(chunk[0])
            return f.read(chunk[1])

    #iterator that splits a chunk into units
    @staticmethod
    def parse(chunk):
        for line in chunk.splitlines():
            yield chunk

In the above, we create the class Chunker which has the class method chunkify as well as the static methods, _EOC, read, and parse. The method chunkify does the actual chunking of a given file, returning an iterator that yields tuples containing a chunk’s start and size. It’s a class method so that it can make use of _EOC (end of chunk) static method, to move the pointer to a suitable location to split the chunks. For the simplest case, this is just the end/start of a newline. The read and parse methods read a given chunk from a file and split it into units (single lines in the simplest case) respectively. We make the non-chunkify methods static so that they can be called without the overhead of creating an instance of the class.

Let’s now create some children of this class for specific types of files. First, one of the most well known file types in bioinformatics, FASTA. Below is an segment of a FASTA file. Each entry has a header line, which begins with a ‘>’, followed by a unique identifier for the sequence. After the header line, one or more lines follow giving the sequence. Sequences may be either protein or nucleic acid sequences, and they may contain gaps and/or alignment characters.


>SEQUENCE_1
MTEITAAMVKELRESTGAGMMDCKNALSETNGDFDKAVQLLREKGLGKAAKKADRLAAEG
LVSVKVSDDFTIAAMRPSYLSYEDLDMTFVENEYKALVAELEKENEERRRLKDPNKPEHK
IPQFASRKQLSDAILKEAEEKIKEELKAQGKPEKIWDNIIPGKMNSFIADNSQLDSKLTL
MGQFYVMDDKKTVEQVIAEKEKEFGGKIKIVEFICFEVGEGLEKKTEDFAAEVAAQL
>SEQUENCE_2
SATVSEINSETDFVAKNDQFIALTKDTTAHIQSNSLQSVEELHSSTINGVKFEEYLKSQI
ATIGENLVVRRFATLKAGANGVVNGYIHTNGRVGVVIAAACDSAEVASKSRDLLRQICMH

And here is the file type specific chunker:


from Bio import SeqIO
from cStringIO import StringIO

class Chunker_FASTA(Chunker):

    @staticmethod
    def _EOC(f):
        l = f.readline() #incomplete line
        p = f.tell()
        l = f.readline()
        while l and l[0] != '>': #find the start of sequence
            p = f.tell()
            l = f.readline()
        f.seek(p) #revert one line

    @staticmethod
    def parse(chunk):
        fh = cStringIO.StringIO(chunk)
        for record in SeqIO.parse(fh,"fasta"):
            yield record
        fh.close()

We update the _EOC method to find when one entry finishes and the next begins by locating “>”, following which we rewind the file handle pointer to the start of that line. We also update the parse method to use fasta parser from the BioPython module, this yielding SeqRecord objects for each entry in the chunk.

For a second slightly harder example, here is one designed to work with output produced by bowtie, an aligner of short reads from NGS data. The format consists of of tab separated columns, with the id of each read located in the first column. Note that a single read can align to multiple locations (up to 8 as default!), hence why the same id appears in multiple lines. A small example section of the output is given below.


SRR014374.1 HWI-EAS355_3_Nick_1_1_464_1416 length=36	+	RDN25-2	2502 GTTTCTTTACTTATTCAATGAAGCGG	IIIIIIIIIIIIIIIIIIIIIIIIII	3	
SRR014374.1 HWI-EAS355_3_Nick_1_1_464_1416 length=36	+	RDN37-2	4460	GTTTCTTTACTTATTCAATGAAGCGG	IIIIIIIIIIIIIIIIIIIIIIIIII	3	
SRR014374.1 HWI-EAS355_3_Nick_1_1_464_1416 length=36	+	RDN25-1	2502	GTTTCTTTACTTATTCAATGAAGCGG	IIIIIIIIIIIIIIIIIIIIIIIIII	3	
SRR014374.1 HWI-EAS355_3_Nick_1_1_464_1416 length=36	+	RDN37-1	4460	GTTTCTTTACTTATTCAATGAAGCGG	IIIIIIIIIIIIIIIIIIIIIIIIII	3	
SRR014374.2 HWI-EAS355_3_Nick_1_1_341_1118 length=36	+	RDN37-2	4460	GTTTCTTTACTTATTCAATGAAGCG	IIIIIIIIIIIIIIIIIIIIIIIII	3	
SRR014374.2 HWI-EAS355_3_Nick_1_1_341_1118 length=36	+	RDN25-1	2502	GTTTCTTTACTTATTCAATGAAGCG	IIIIIIIIIIIIIIIIIIIIIIIII	3	
SRR014374.2 HWI-EAS355_3_Nick_1_1_341_1118 length=36	+	RDN37-1	4460	GTTTCTTTACTTATTCAATGAAGCG	IIIIIIIIIIIIIIIIIIIIIIIII	3	
SRR014374.2 HWI-EAS355_3_Nick_1_1_341_1118 length=36	+	RDN25-2	2502	GTTTCTTTACTTATTCAATGAAGCG	IIIIIIIIIIIIIIIIIIIIIIIII	3	
SRR014374.8 HWI-EAS355_3_Nick_1_1_187_1549 length=36	+	RDN25-2	2502	GTTTCTTTACTTATTCAATGAAGCGG	IIIIIIIIIIIIIIIIIIIIIIIIII	3	
SRR014374.8 HWI-EAS355_3_Nick_1_1_187_1549 length=36	+	RDN37-2	4460	GTTTCTTTACTTATTCAATGAAGCGG	IIIIIIIIIIIIIIIIIIIIIIIIII	3

with the corresponding chunker given by:


class Chunker_BWT(chunky.Chunker):

    @staticmethod
    def _EOC(f):
        l = f.readline()#incomplete line
        l = f.readline()
        if not l: return #EOF
        readID = l.split()[0]
        while l and (l.split()[0] != readID): #Keep reading lines until read IDs don't match
            p = f.tell()	
            l = f.readline()
        f.seek(p) #revert one line

    @staticmethod
    def parse(chunk):
        lines = chunk.splitlines()
        N = len(lines)
        i = 0 
        while i < N:
            readID = lines[i].split('\t')[0]
            j = i
            while lines[j].split('\t')[0] == readID:
                j += 1
                if j == N:
                    break
            yield lines[i:j]
            i = j

This time, the end of chunk is located by reading lines until there is a switch in the read id, whereupon we revert one line. For parsing, we yield all the different locations a given read aligns to as a single entry.

Hopefully these examples show you how the parent class can be expanded upon easily for most file types. Let’s now combine these various chunkers with the code from previous post to show how we can enable multicore parallel processing of the chunks they yield. The code below contains a few generalised wrapper functions which work in tandem with any of the above chunkers to allow most tasks to be parallelised .


import multiprocessing as mp, sys

def _printMP(text):
    print text
    sys.stdout.flush()

def _workerMP(chunk,inFile,jobID,worker,kwargs):
    _printMP("Processing chunk: "+str(jobID))
    output = worker(chunk,inFile,**kwargs)
    _printMP("Finished chunk: "+str(jobID))
    return output	

def main(inFile,worker,chunker=Chunker,cores=1,kwargs={}):
    pool = mp.Pool(cores)

    jobs = []
    for jobID,chunk in enumerate(chunker.chunkify(inFile)):
        job = pool.apply_async(_workerMP,(chunk,inFile,jobID,worker,kwargs))
        jobs.append(job)

    output = []
    for job in jobs:
        output.append( job.get() )

    pool.close()
    
    return output

The main function should be recognisable as the code from the previous post. It generates the pool of workers, ideally one for each core, before using the given chunker to split the corresponding file into a series of chunks for processing. Unlike before, we collect the output given by each job and return it after processing is finished. The main function acts as wrapper allowing us to specify different processing functions and different chunkers, as given by the variables worker and chunker respectively. We have wrapped the processing function call within the function _workerMP which prints to the terminal as tasks are completed. It uses the function _printMP to do this, as you need to flush the terminal after a print statement when using multi core processing, otherwise nothing appears until all tasks are completed.

Let’s finish by showing an example of how we would use the above to do the same task as we did at the start of this post, counting the sequences within a fasta file, using the base chunker:


def seq_counter(chunk,inFile):
    data = Chunker.read(inFile,chunk)
    return data.count('>')

and using the FASTA chunker:


def seq_counter_2(chunk,inFile):
    data = list(Chunker_FASTA.parse(Chunker_FASTA.read(inFile,chunk)))
    return len(data)

And time they take to count the sequences within the 14GB file from before:


>>> os.path.getsize("SRR951828.fa")
14944287128
>>> x=time.time();f = open("SRR951828.fa");sum([l.count(">") for l in f]);time.time()-x
136829250
190.05533599853516
>>> x=time.time();f = open("SRR951828.fa");sum([c.count(">") for c in iter(lambda: f.read(1024*1024),"")]);time.time()-x
136829250
26.343637943267822
>>> x=time.time();sum(main("SRR951828.fa",seq_counter,cores=8));time.time()-x
136829250
4.36846399307251
>>> x=time.time();main("SRR951828.fa",seq_counter_2,Chunker_FASTA,8);time.time()-x
136829250
398.94060492515564

Let’s ignore that last one, as the slowdown is due to turning the entries into BioPython SeqRecords. The prior one, which combines chunking and multicore processing, has roughly a factor of 50 speed up. I’m sure this could be further reduced using more cores and/or optimising the chunk size, however, this difference alone can change something from being computationally implausible, to plausible. Not bad for only a few lines of code.

Anyway, as before, I hope that some of the above was either new or even perhaps helpful to you. If you know of a better way to do things (in python), then I’d be very interested to hear about it. If I feel like it, I may follow this up with a post about how to integrate a queue into the above which outputs the result of each job as they are produced. In the above, we currently hold collate all the results in the memory, which has the potential to cause a memory overflow depending on what is being returned.

Processing large files using python

In the last year or so, and with my increased focus on ribo-seq data, I have come to fully appreciate what the term big data means. The ribo-seq studies in their raw forms can easily reach into hundreds of GBs, which means that processing them in both a timely and efficient manner requires some thought. In this blog post, and hopefully those following, I want to detail some of the methods I have come up (read: pieced together from multiple stack exchange posts), that help me take on data of this magnitude. Specifically I will be detailing methods for python and R, though some of the methods are transferrable to other languages.

My first big data tip for python is learning how to break your files into smaller units (or chunks) in a manner that you can make use of multiple processors. Let’s start with the simplest way to read a file in python.


with open("input.txt") as f:
    data = f.readlines()
    for line in data:
        process(line)

This mistake made above, with regards to big data, is that it reads all the data into RAM before attempting to process it line by line. This is likely the simplest way to cause the memory to overflow and an error raised. Let’s fix this by reading the data in line by line, so that only a single line is stored in the RAM at any given time.


with open("input.txt") as f:
    for line in f:
        process(line)

This is a big improvement, namely it doesn’t crash when fed a big file (though also it’s shorter!). Next we should attempt to speed this up a bit by making use of all these otherwise idle cores.


import multiprocessing as mp

#init objects
pool = mp.Pool(cores)
jobs = []

#create jobs
with open("input.txt") as f:
    for line in f:
        jobs.append( pool.apply_async(process,(line)) )

#wait for all jobs to finish
for job in jobs:
    job.get()

#clean up
pool.close()

Provided the order of which you process the lines don’t matter, the above generates a set (pool) of workers, ideally one for each core, before creating a bunch of tasks (jobs), one for each line, for the workers to do. I tend to use the Pool object provided by the multiprocessing module due to ease of use, however, you can spawn and control individual workers using mp.Process if you want finer control. For mere number crunching, the Pool object is very good.

While the above is now making use of all those cores, it sadly runs into memory problems once again. We specifically use apply_async function so that the pool isn’t blocked while each line processes. However, in doing so, all the data is read into memory once again; this time stored as individual lines associated with each job, waiting inline to be processed. As such, the memory will again overflow. Ideally the method will only read the line into memory when it is its turn to be processed.


import multiprocessing as mp

def process_wrapper(lineID):
    with open("input.txt") as f:
        for i,line in enumerate(f):
            if i != lineID:
                continue
            else:
                process(line)
                break

#init objects
pool = mp.Pool(cores)
jobs = []

#create jobs
with open("input.txt") as f:
    for ID,line in enumerate(f):
        jobs.append( pool.apply_async(process_wrapper,(ID)) )

#wait for all jobs to finish
for job in jobs:
    job.get()

#clean up
pool.close()

Above we’ve now changed the function fed to pool of workers to include opening the file, locating the specified line, reading it into memory, and then processing it. The only input now stored for each job spawned is the line number, thereby preventing the memory overflow. Sadly, the overhead involved in having to locate the line by reading iteratively through the file for each job is untenable, getting progressively more time consuming as you get further into the file. To avoid this we can use the seek function of file objects which skips you to a particular location within a file. Combining with the tell function, which returns the current location within a file, gives:


import multiprocessing as mp

def process_wrapper(lineByte):
    with open("input.txt") as f:
        f.seek(lineByte)
        line = f.readline()
        process(line)

#init objects
pool = mp.Pool(cores)
jobs = []

#create jobs
with open("input.txt") as f:
    nextLineByte = f.tell()
    for line in f:
        jobs.append( pool.apply_async(process_wrapper,(nextLineByte)) )
        nextLineByte = f.tell()

#wait for all jobs to finish
for job in jobs:
    job.get()

#clean up
pool.close()

Using seek we can move directly to the correct part of the file, whereupon we read a line into the memory and process it. We have to be careful to correctly handle the first and last lines, but otherwise this does exactly what we set out, namely using all the cores to process a given file while not overflowing the memory.

I’ll finish this post with a slight upgrade to the above as there is a reasonable amount of overhead associated with opening and closing the file for each individual line. If we process multiple lines of the file at a time as a chunk, we can reduce these operations. The biggest technicality when doing this is noting that when you jump to a location in a file, you are likely not located at the start of a line. For a simple file, as in this example, this just means you need to call readline, which reads to next newline character. More complex file types likely require additional code to locate a suitable location to start/end a chunk.


import multiprocessing as mp,os

def process_wrapper(chunkStart, chunkSize):
    with open("input.txt") as f:
        f.seek(chunkStart)
        lines = f.read(chunkSize).splitlines()
        for line in lines:
            process(line)

def chunkify(fname,size=1024*1024):
    fileEnd = os.path.getsize(fname)
    with open(fname,'r') as f:
        chunkEnd = f.tell()
    while True:
        chunkStart = chunkEnd
        f.seek(size,1)
        f.readline()
        chunkEnd = f.tell()
        yield chunkStart, chunkEnd - chunkStart
        if chunkEnd > fileEnd:
            break

#init objects
pool = mp.Pool(cores)
jobs = []

#create jobs
for chunkStart,chunkSize in chunkify("input.txt"):
    jobs.append( pool.apply_async(process_wrapper,(chunkStart,chunkSize)) )

#wait for all jobs to finish
for job in jobs:
    job.get()

#clean up
pool.close()

Anyway, I hope that some of the above was either new or even perhaps helpful to you. If you know of a better way to do things (in python), then I’d be very interested to hear about it. In another post coming in the near future, I will expanded on this code, turning it into a parent class from which create multiple children to use with various file types.

Colour wisely…

Colour – the attribute of an image that makes it acceptable or destined for the bin. Colour has a funny effect on us – it’s a double-edged sword that greatly strengthens, or weakens data representation in such a huge level. No one really talks about what’s a good way to colour an image or a graph, but it’s something that most can agree as being pleasing, or disgusting. There are two distinctive advantages to colouring a graph: it conveys both quantitative and categorical information very, very well. Thus, I will provide a brief overview (with code) on how colour can be used to display both quantitative and qualitative information. (*On the note of colours, Nick has previously discussed how colourblindness must be considered in visualising data…).

1. Colour conveys quantitative information.
A huge advantage of colour is that it can provide quantitative information, but this has to be done correctly. Here are three graphs showing the exact same information (the joint density of two normal distributions) and  we can see from the get-go which method is the best at representing the density of the two normal distributions:

Colouring the same graph using three different colour maps.

Colouring the same graph using three different colour maps.

If you thought the middle one was the best one, I’d agree too. Why would I say that, despite it being grayscale and seemingly being the least colourful of them all?

  • Colour is not limited to hues (i.e. whether it’s red/white/blue/green etc. etc.); ‘colour’ is also achieved by saturation and brightness (i.e., how vivid a colour is, or dark/light it is). In the case of the middle graph, we’re using brightness to indicate the variations in density and this is a more intuitive display of variations in density. Another advantage of using shades as the means to portray colour is that it will most likely be OK with colourblind users.
  • Why does the graph on the right not work for this example? This is a case where we use a “sequential” colour map to convey the differences in density. Although the colour legend clarifies what colour belongs to which density bin, without it, it’s very difficult to tell what “red” is with respect to “yellow”. Obviously by having a colour map we know that red means high density and yellow is lower, but without the legend, we can interpret the colours very differently, e.g. as categories, rather than quantities. Basically, when you decide on a sequential colour map, its use must be handled well, and a colour map/legend is critical. Otherwise, we risk putting colours as categories, rather than as continuous values.
  • Why is the left graph not working well? This is an example of a “diverging” colourmap.
    It’s somewhat clear that blue and red define two distinct quantities. Despite this, a major error of this colour map comes in the fact that there’s a white colour in the middle. If the white was used as a “zero crossing” — basically, where a white means the value is 0 — the diverging colour map would have been a more effective tool. However, we can see that matplotlib used white as the median value (by default); this sadly creates the false illusion of a 0 value, as our eyes tend to associate white with missing data, or ‘blanks’. Even if this isn’t your biggest beef with the divergent colour map, we run into the same colour as the sequential colour map, where blue and red don’t convey information (unless specified), and the darkness/lightness of the blue and red are not linked very well without the white in the middle. Thus, it doesn’t do either job very well in this graph. Basically, avoid using divergent colourmaps unless we have two different quantities of values (e.g. data spans from -1 to +1).

2. Colour displays categorical information.
An obvious use of colour is the ability to categorise our data. Anything as simple as a line chart with multiple lines will tell you that colour is terrific at distinguishing groups. This time, notice that the different colour schemes have very different effects:

Colour schemes can instantly differentiate groups.

Colour schemes can instantly differentiate groups.

Notice how this time around, the greyscale method (right) was clearly the losing choice. To begin with, it’s hard to pick out what’s the difference between persons A,B,C, but there’s almost a temptation to think that person A morphs into person C! However, on the left, with a distinct set of colours, there is a clear distinction of persons A, B, and C as the three separate colours. Although a set of distinct three colours is a good thing, bear in mind the following…

  • Make sure the colours don’t clash with respect to lightness! Try to pick something that’s distinct (blue/red/green), rather than two colours which can be interpreted as two shades of the same colour (red/pink, blue/cyan, etc.)
  • Pick a palette to choose from – a rainbow is typically the best choice just because it’s the most natural, but feel free to choose your own set of hues. Also include white and black as necessary, so long as it’s clear that they are also part of the palette. White in particular would only work if you have a black outline.
  • Keep in mind that colour blind readers can have trouble with certain colour combinations (red/yellow/green) and it’s best to steer toward colourblind-friendly palettes.
import numpy as np
import matplotlib.pyplot as plt
import scipy.stats as sp
from mpl_toolkits.axes_grid1 import make_axes_locatable

### Part 1
# Sample 250 points
np.random.seed(30)
x = np.random.normal(size = 250)
np.random.seed(71)
y = np.random.normal(size = 250)

# Assume the limits of a standard normal are at -3, 3
pmin, pmax = -3, 3

# Create a meshgrid that is 250x250
xgrid, ygrid = np.mgrid[pmin:pmax:250j, pmin:pmax:250j]
pts = np.vstack([xgrid.ravel(), ygrid.ravel()]) # ravel unwinds xgrid from a 250x250 matrix into a 62500x1 array

data = np.vstack([x,y])
kernel = sp.gaussian_kde(data)
density = np.reshape(kernel(pts).T, xgrid.shape) # Evaluate the density for each point in pts, then reshape back to a 250x250 matrix

greys = plt.cm.Greys
bwr = plt.cm.bwr
jet = plt.cm.jet

# Create 3 contour plots
fig, ax = plt.subplots(1,3)
g0 = ax[0].contourf(xgrid, ygrid, density, cmap = bwr)
c0 = ax[0].contour(xgrid, ygrid, density, colors = 'k') # Create contour lines, all black
g1 = ax[1].contourf(xgrid, ygrid, density, cmap = greys)
c1 = ax[1].contour(xgrid, ygrid, density, colors = 'k') # Create contour lines, all black
g2 = ax[2].contourf(xgrid, ygrid, density, cmap = jet)
c2 = ax[2].contour(xgrid, ygrid, density, colors = 'k') # Create contour lines, all black

# Divide each axis then place a colourbar next to it
div0 = make_axes_locatable(ax[0])
cax0 = div0.append_axes('right', size = '10%', pad = 0.1) # Append a new axes object
cb0  = plt.colorbar(g0, cax = cax0)

div1 = make_axes_locatable(ax[1])
cax1 = div1.append_axes('right', size = '10%', pad = 0.1)
cb1  = plt.colorbar(g1, cax = cax1)

div2 = make_axes_locatable(ax[2])
cax2 = div2.append_axes('right', size = '10%', pad = 0.1)
cb2  = plt.colorbar(g2, cax = cax2)

fig.set_size_inches((15,5))
plt.tight_layout()
plt.savefig('normals.png', dpi = 300)
plt.close('all')

### Part 2
years = np.arange(1999, 2017, 1)
np.random.seed(20)
progress1 = np.random.randint(low=500, high =600, size = len(years))
np.random.seed(30)
progress2 = np.random.randint(low=500, high =600, size = len(years))
np.random.seed(40)
progress3 = np.random.randint(low=500, high =600, size = len(years))

fig, ax = plt.subplots(1,2)
ax[0].plot(years, progress1, label = 'Person A', c = '#348ABD')
ax[0].plot(years, progress2, label = 'Person B', c = '#00de00')
ax[0].plot(years, progress3, label = 'Person C', c = '#A60628')
ax[0].set_xlabel("Years")
ax[0].set_ylabel("Progress")
ax[0].legend()

ax[1].plot(years, progress1, label = 'Person A', c = 'black')
ax[1].plot(years, progress2, label = 'Person B', c = 'gray')
ax[1].plot(years, progress3, label = 'Person C', c = '#3c3c3c')
ax[1].set_xlabel("Years")
ax[1].set_ylabel("Progress")
ax[1].legend()

fig.set_size_inches((10,5))
plt.tight_layout()
plt.savefig('colourgrps.png', dpi = 300)
plt.close('all')

Seventh Joint Sheffield Conference on Cheminformatics Part 1 (#ShefChem16)

In early July I attended the the Seventh Joint Sheffield Conference on Cheminformatics. There was a variety of talks with speakers at all stages of their career. I was lucky enough to be invited to speak at the conference, and gave my first conference talk! I have written two blog posts about the conference: part 1 briefly describes a talk that I found interesting and part 2 describes the work I spoke about at the conference.

One of the most interesting parts of the conference was the active twitter presence. #ShefChem16. All of the talks were live tweeted which provided a summary of each talk and also included links to software or references. It also allowed speakers to gain insight and feedback on their talk instantly.

One of the talks I found most interesting presented the Protein-Ligand Interaction Profiler (PLIP). It is a method for the detection of protein-ligand interactions. PLIP is open-source and has a web-based online tool and a command-line tool. Unlike PyMol which only calculates polar contacts, and not the type of interaction, PLIP calculates 8 different types of interactions: hydrogen bonding, hydrophobic, π-π stacking, π-cation interactions, salt bridges, water bridges, halogen bonds, metal complexes. For a given pdb file the interactions are calculated and shown in a publication quality figure shown here.

Screen Shot 2016-07-20 at 14.16.23

The display can also be downloaded as a PyMol session so the display can be modified. 

This tool is an extremely useful way to calculate protein-ligand interactions and can be used to find the types of interactions formed by the protein-ligand complex.

PLIP can be found here: https://projects.biotec.tu-dresden.de/plip-web/plip/

Tracked changes in LaTeX

Maybe people keep telling you Word is great but you are just too emotionally attached to LaTeX to consider using anything else. It just looks so beautiful. Besides, you would have to leave your beloved linux environment (maybe that’s just me), so you stick with what you know. You work for many weeks long and hard, finally producing a draft of a paper that gets the all clear from your supervisor to submit to journal X. Eventually you hear back and the reviewers have responded with some good ideas and a few pedantic points. Apparently this time the journal wants a tracked changes version to go with your revised manuscript.

Highlighting every change sounds like a lot of bother, and besides, you’d have to process the highlighted version to generate the clean version they want you to submit alongside it. There must be a better way, and one that doesn’t involve converting your document to Word.

Thankfully, the internet has an answer! Check out this little package changes which will do just what you need. As long as you annotate using \deleted{}, \replaced{} and \added{} along the way, you will have to change just one word of your tex source file in order to produce the highlighted and final versions. It even comes with a handy bash script to get rid of the resulting mess when you’re happy with the result, leaving you with a clean final tex source file.

Screenshot from 2016-07-12 19-45-12 Screenshot from 2016-07-12 19-43-34 Screenshot from 2016-07-12 19-44-53 Screenshot from 2016-07-12 19-44-19

The die-hard Word fans won’t be impressed, but you will be very satisfied that you have found a nice little solution that does just the job you want it to. It’s actually capable of much more, including comments by multiple authors, customisation of colours and styles, and an automatically generated summary of changes. I have heard good things about ShareLaTeX for collaboration, but this simple package will get you a long way if you are not keen to start paying money yet.