Google Summer of Code

(Difference between revisions)
Jump to: navigation, search
(Synergy from Biopython and Mocapy++)
(Reorg: we have two GSoC 2013 students, Zheng and Yanbo)
(39 intermediate revisions by 6 users not shown)
Line 1: Line 1:
As part of the Open Bioinformatics Foundation, Biopython is participating in Google Summer of Code (GSoC) again in 2010. We are supporting João Rodrigues in his project, "[[GSOC2010_Joao|Extending Bio.PDB: broadening the usefulness of BioPython's Structural Biology module]]."
+
== Introduction ==
  
In 2009, Biopython was involved with GSoC in collaboration with our friends at [https://www.nescent.org/wg_phyloinformatics/Main_Page NESCent], and had two projects funded:
+
For the past several years, Biopython-related GSoC projects have been successfully run under
 +
the mentorship of the [http://www.open-bio.org/wiki/Google_Summer_of_Code Open Bioinformatics Foundation (OBF)] and the
 +
[http://informatics.nescent.org/wiki/Phyloinformatics_Summer_of_Code_2013 National Evolutionary Synthesis Center (NESCent)].
  
* Nick Matzke worked on [https://www.nescent.org/wg_phyloinformatics/Phyloinformatics_Summer_of_Code_2009#Biogeographical_Phylogenetics_for_BioPython Biogeographical Phylogenetics].
+
Please read those organizations' GSoC pages and the [http://code.google.com/soc main Google Summer of Code page] for more details about the program.
* Eric Talevich added support for [https://www.nescent.org/wg_phyloinformatics/Phyloinformatics_Summer_of_Code_2009#Biopython_support_for_parsing_and_writing_phyloXML parsing and writing phyloXML].
+
  
In 2010, another project was funded:
+
== Mentor List ==
  
* João Rodrigues worked on [http://www.biopython.org/wiki/GSOC2010_Joao the Structural Biology module Bio.PDB] adding several features used in everyday structural bioinformatics.
+
Usually, each BioPython proposal has one or more mentors assigned to it. Nevertheless, we encourage potential students/mentors to contact the [http://biopython.org/wiki/Mailing_lists mailing lists] with their own ideas for proposals. There is therefore not a set list of 'available' mentors, since it highly depends on which projects are proposed every year.
  
Please read the [http://www.open-bio.org/wiki/Google_Summer_of_Code GSoC page at the Open Bioinformatics Foundation] and the main [http://code.google.com/soc Google Summer of Code] page for more details about the program. If you are interested in contributing as a mentor or student next year, please introduce yourself on the [http://biopython.org/wiki/Mailing_lists mailing list].
+
Past mentors include:
  
== 2011 Project ideas ==
+
*  [http://casbon.me/ James Casbon]
 +
*  [https://github.com/chapmanb Brad Chapman]
 +
*  [http://www.hutton.ac.uk/staff/peter-cock Peter Cock]
 +
*  [http://wiki.binf.ku.dk/User:Thomas_Hamelryck Thomas Hamelryck]
 +
*  [http://www.linkedin.com/in/reece Reece Hart]
 +
*  [http://nmr.chem.uu.nl/~joao João Rodrigues]
 +
*  [http://etal.myweb.uga.edu/ Eric Talevich]
  
=== Biopython and PyCogent interoperability ===
 
  
; Rationale : [http://pycogent.sourceforge.net/ PyCogent] and [http://biopython.org/wiki/Main_Page Biopython] are two widely used toolkits for performing computational biology and bioinformatics work in Python. The libraries have had traditionally different focuses: with Biopython focusing on sequence parsing and retrieval and PyCogent on evolutionary and phylogenetic processing. Both user communities would benefit from increased interoperability between the code bases, easing the developing of complex workflows.
+
== Current projects ==
  
; Approach : The student would focus on soliciting use case scenarios from developers and the larger communities associated with both projects, and use these as the basis for adding glue code and documentation to both libraries. Some use cases of immediate interest as a starting point are:
+
This year, NESCent was selected as a mentoring organization for Google Summer of Code. See also [http://informatics.nescent.org/wiki/Phyloinformatics_Summer_of_Code_2013 NESCent's GSoC page].
  
:* Allow round-trip conversion between biopython and pycogent core objects (sequence, alignment, tree, etc.).
+
===  Codon alignment and analysis ===
:* Building workflows using Codon Usage analyses in PyCogent with clustering code in Biopython.
+
;  Student
:* Connecting Biopython acquired sequences to PyCogent's alignment, phylogenetic tree preparation and tree visualization code.
+
: Zheng Ruan
:* Integrate Biopython's [http://biopython.org/wiki/Phylo phyloXML support], developed during GSoC 2009, with PyCogent.
+
;  Rationale
:* Develop a standardised controller architecture for interrogation of genome databases by extending PyCogent's Ensembl code, including export to Biopython objects.
+
:  A codon alignment is an alignment of nucleotide sequences in which the trinucleotides correspond directly to amino acids in the translated protein product. This carries important information which can be used for several analysis, notably analysis of selection pressures. This project extends Biopython to support this data type and these analyses.
 +
;  Approach & Goals
 +
: Useful features include:
 +
* Conversion of a set of unaligned nucleic acid sequences and a corresponding protein sequence alignment to a codon alignment.
 +
* Calculation of selection pressure from the ratio of nonsynonomous to synonomous site replacements, and related functions.
 +
* Model selection.
 +
* Possible extension of [[AlignIO]] and the MultipleSeqAlignment class to take full advantage of codon alignments, including validation (testing for frame shifts, etc.)
 +
;  Difficulty and needed skills
 +
: Medium, depending on ambition. Familiarity with the Biopython's existing alignment classes and functions, or equivalents in BioPerl, BioJava or BioRuby (e.g.), will be helpful. Understanding of the practical uses of codon alignments, or at least a basic understanding of molecular biology, is important. Some basic math is involved, essentially reading a few equations and converting them to code. One useful book to have on hand is ''[http://abacus.gene.ucl.ac.uk/CME/ Computational Molecular Evolution]'' by Ziheng Yang.
 +
;  Mentors
 +
: [http://etal.myweb.uga.edu/ Eric Talevich], [https://github.com/peterjc/ Peter Cock]
  
; Challenges : This project provides the student with a lot of freedom to create useful interoperability between two feature rich libraries. As opposed to projects which might require churning out more lines of code, the major challenge here will be defining useful APIs and interfaces for existing code. High level inventiveness and coding skill will be required for generating glue code; we feel library integration is an extremely beneficial skill. We also value clear use case based documentation to support the new interfaces.
+
===  Bio.Phylo: filling in the gaps ===
 +
; Student
 +
: Yanbo Ye
 +
;  Rationale
 +
:  While the [[Phylo]] module in Biopython supports I/O and basic tree operations, there are some important components that remain to be implemented to better support phylogenetic workflows.
 +
;  Approach & Goals
 +
: This "idea" is intentially left open-ended -- some potentially useful features are:
 +
* A "Phylo.consensus" module with functions for the consensus of multiple trees. E.g.: Strict consensus, as Bio.Nexus already implements; perhaps other methods like Adams consensus.
 +
* A function to calculate bootstrap support for branches given a target or "master" tree and a series of bootstrap replicate trees (usually read incrementally from a Newick file).
 +
* Functions for comparison of two or more trees.
 +
* Simple algorithms for tree inference, i.e. neighbor-joining and parsimony tree estimation. For small alignments (and perhaps medium-sized ones with PyPy), it would be nice to run these without an external program, e.g. to construct a guide tree for another algorithm or quickly view a phylogenetic clustering of sequences.
 +
* Tree visualizations: A proper draw_unrooted function to perform radial layout, with an optional "iterations" argument to use Felsenstein's Equal Daylight algorithm. Circular radial diagrams are also popular these days. Any new function should support the same arguments as the existing Phylo.draw.
 +
;  Difficulty and needed skills
 +
:  Medium-to-advanced programming skill in Python -- it's important for these implementations to be reasonably efficient, though we don't aim to compete with the fastest stand-alone implementations of these algorithms. Knowledge of phylogenetic methods is critical; for reference, you might like to have a copy of Joe Felsenstein's ''[http://www.sinauer.com/detail.php?id=1775 Inferring Phylogenies]''. Tree visualizations are done with matplotlib.
 +
;  Mentors
 +
:  [http://etal.myweb.uga.edu/ Eric Talevich], others welcome
  
; Involved toolkits or projects :
 
  
:* [http://biopython.org/wiki/Main_Page Biopython]
+
== Past Proposals ==
:* [http://pycogent.sourceforge.net/ PyCogent]
+
  
; Degree of difficulty and needed skills : Medium to Hard. At a minimum, the student will need to be highly competent in Python and become familiar with core objects in PyCogent and Biopython. Sub-projects will require additional expertise, for instance: familiarity with concepts in phylogenetics and genome biology; understanding SQL dialects.
+
=== 2012 ===
 +
==== [http://biopython.org/wiki/SearchIO SearchIO] ====
 +
; Rationale
 +
:  Biopython has general APIs for parsing and writing assorted sequence file formats (SeqIO), multiple sequence alignments (AlignIO), phylogenetic trees (Phylo) and motifs (Bio.Motif). An obvious omission is something equivalent to BioPerl's SearchIO. The goal of this proposal is to develop an easy-to-use Python interface in the same style as SeqIO, AlignIO, etc but for pairwise search results. This would aim to cover EMBOSS muscle & water, BLAST XML, BLAST tabular, HMMER, Bill Pearson's FASTA alignments, and so on.
 +
;  Approach
 +
: Much of the low level parsing code to handle these file formats already exists in Biopython, and much as the SeqIO and AlignIO modules are linked and share code, similar links apply to the proposed SearchIO module when using pairwise alignment file formats. However, SearchIO will also support pairwise search results where the pairwise sequence alignment itself is not available (e.g. the default BLAST tabular output). A crucial aspect of this work will be to design a pairwise-search-result object heirachy that reflects this, probably with a subclass inheriting from both the pairwise-search-result and the existing MultipleSequenceAlignment object. Beyond the initial challenge of an iterator based parsing and writing framework, random access akin to the Bio.SeqIO.index and index_db functionality would be most desirable for working with large datasets.
 +
; Challenges
 +
: The project will cover a range of important file formats from major Bioinformatics tools, thus will require familiarity with running these tools, and understanding their output and its meaning. Inter-converting file formats is part of this.
 +
;  Difficulty and needed skills
 +
:  Medium/Hard depending on how many objectives are attempted. The student needs to be fluent in Python and have knowledge of the BioPython codebase. Experience with all of the command line tools listed would be clear advantages, as would first hand experience using BioPerl's SearchIO. You will also need to know or learn the git version control system.
 +
; Mentors
 +
:  [http://www.hutton.ac.uk/staff/peter-cock Peter Cock]
  
; Mentors : [http://jcsmr.anu.edu.au/org/dmb/compgen/ Gavin Huttley], [http://chem.colorado.edu/index.php?option=com_content&view=article&id=263:rob-knight Rob Knight], [http://bcbio.wordpress.com Brad Chapman], [[User:EricTalevich|Eric Talevich]]
+
====  [http://arklenna.tumblr.com/tagged/gsoc2012 Representation and manipulation of genomic variants] ====
 +
;  Rationale
 +
:  Computational analysis of genomic variation requires the ability to reliably communicate and manipulate variants. The goal of this project is to provide facilities within BioPython to represent sequence variation objects, convert them to and from common human and file representations, and provide common manipulations on them.
 +
;  Approach & Goals
 +
* Object representation
 +
** identify variation types to be represented (SNV, CNV, repeats, inversions, etc)
 +
** develop internal machine representation for variation types
 +
** ensure coverage of essential standards, including HGVS, GFF, VCF
 +
* External representations
 +
** write parser and generators between objects and external string and file formats
 +
* Manipulations
 +
** canonicalize variations with more than one valid representation (e.g., ins versus dup and left shifting repeats).
 +
** develop coordinate mapping between genomic, cDNA, and protein sequences (HGVS)
 +
* Other
 +
** release code to appropriate community efforts and write short manuscript
 +
** implement web service for HGVS conversion
 +
;  Difficulty and needed skills
 +
:  Easy-to-Medium depending on how many objectives are attempted. The student will need have skills in most or all of: basic molecular biology (genomes, transcripts, proteins), genomic variation, Python, BioPython, Perl, BioPerl, NCBI Eutilities and/or Ensembl API. Experience with computer grammars is highly desirable. You will also need to know or learn the git version control system.
 +
; Mentors
 +
: [http://www.linkedin.com/in/reece Reece Hart]
 +
:  [https://github.com/chapmanb Brad Chapman]  
 +
[http://casbon.me/ James Casbon]
 +
=== 2011 ===
 +
====  [http://biopython.org/wiki/GSoC2011_mtrellet Biomolecular Interface Analysis] ====
 +
;  Student
 +
: Mikael Trellet
 +
;  Rationale
 +
: Analysis of protein-protein complexes interfaces at a residue level yields significant information on the overall binding process. Such information can be broadly used for example in binding affinity studies, interface design, and enzymology. To tap into it, there is a need for tools that systematically and automatically analyze protein structures, or that provide means to this end. Protorop (http://www.bioinformatics.sussex.ac.uk/protorp/) is an example of such a tool and the elevated number of citations the server has had since its publication acknowledge its importance. However, being a webserver, Protorop is not suited for large-scale analysis and it leaves the community dependent on its maintainers to keep the service available. On the other hand, Biopython’s structural biology module, Bio.PDB, provides the ideal parsing machinery and programmatic structures for the development of an offline, open-source library for interface analysis. Such a library could be easily used in large-scale analysis of protein-protein interfaces, for example in the CAPRI experiment evaluation or in benchmark statistics. It would be also reasonable, if time permits, to extend this module to deal with protein-DNA or protein-RNA complexes, as Biopython supports nucleic acids already.
 +
;  Approach & Goals
 +
* Add the new module backbone in current Bio.PDB code base
 +
** Evaluate possible code reuse and call it into the new module
 +
** Try simple calculations to be sure that there is stability between the different modules (parsing for example) and functions
 +
* Define a stable benchmark
 +
** Select few PDB files among interface size and proteins size would be different
 +
* Extend IUPAC.Data module with residue information
 +
** Deduce residues weight from Atom instead of direct dictionary storage
 +
** Polar/charge character (dictionary or influenced by pH)
 +
** Hydrophobicity scale(s)
 +
* Implement Extended Residue class as a subclass of Residue
 +
* Implement Interface object and InterfaceAnalysis module
 +
* Develop functions for interface analysis
 +
** Calculation of interface polar character statistics (% of polar residues, apolar, etc)
 +
** Calculation of BSA calling MSMS or HSA
 +
** Calculation of SS element statistics in the interface through DSSP
 +
** Unit tests and use of results as input for further calculations by other tools and scripts
 +
* Develop functions for Interface comparison
 +
* Code organization and final testing
  
=== Accessing R phylogenetic tools from Python ===
+
;  Difficulty and needed skills
 +
:  Easy/Medium. Working knowledge of the Bio.PDB module of BioPython. Knowledge of structural biology in general and associated file formats (PDB).
 +
;  Mentors
 +
:  [http://nmr.chem.uu.nl/~joao João Rodrigues]
 +
:  [http://etal.myweb.uga.edu/ Eric Talevich]
  
; Rationale : The [http://www.r-project.org/ R statistical language] is a powerful open-source environment for statistical computation and visualization. [http://www.python.org/ Python] serves as an excellent complement to R since it has a wide variety of available libraries to make data processing, analysis, and web presentation easier. The two can be smoothly interfaced using [http://bitbucket.org/lgautier/rpy2/ Rpy2], allowing programmers to leverage the best features of each language. Here we propose to build Rpy2 library components to help ease access to phylogenetic and biogeographical libraries in R.
+
====  [http://biopython.org/wiki/GSOC2011_Mocapy A Python bridge for Mocapy++] ====
 +
;  Student
 +
: Michele Silva
 +
;  Rationale
 +
: Discovering the structure of biomolecules is one of the biggest problems in biology. Given an amino acid or base sequence, what is the three dimensional structure? One approach to biomolecular structure prediction is the construction of probabilistic models. A Bayesian network is a probabilistic model composed of a set of variables and their joint probability distribution, represented as a directed acyclic graph. A dynamic Bayesian network is a Bayesian network that represents sequences of variables. These sequences can be time-series or sequences of symbols, such as protein sequences. Directional statistics is concerned mainly with observations which are unit vectors in the plane or in three-dimensional space. The sample space is typically a circle or a sphere. There must be special directional methods which take into account the structure of the sample spaces. The union of graphical models and directional statistics allows the development of probabilistic models of biomolecular structures. Through the use of dynamic Bayesian networks with directional output it becomes possible to construct a joint probability distribution over sequence and structure. Biomolecular structures can be represented in a geometrically natural, continuous space. Mocapy++ is an open source toolkit for inference and learning using dynamic Bayesian networks that provides support for directional statistics. Mocapy++ is excellent for constructing probabilistic models of biomolecular structures; it has been used to develop models of protein and RNA structure in atomic detail. Mocapy++ is used in several high-impact publications, and will form the core of the molecular modeling package Phaistos, which will be released soon. The goal of this project is to develop a highly useful Python interface to Mocapy++, and to integrate that interface with the Biopython project. Through the Bio.PDB module, Biopython provides excellent functionality for data mining biomolecular structure databases. Integrating Mocapy++ and Biopython will allow training a probabilistic model using data extracted from a database. Integrating Mocapy++ with Biopython will create a powerful toolkit for researchers to quickly implement and test new ideas, try a variety of approaches and refine their methods. It will provide strong support for the field of biomolecular structure prediction, design, and simulation.
 +
;  Approach & Goals
 +
: Mocapy++ is a machine learning toolkit for training and using Bayesian networks. It has been used to develop probabilistic models of biomolecular structures. The goal of this project is to develop a Python interface to Mocapy++ and integrate it with Biopython. This will allow the training of a probabilistic model using data extracted from a database. The integration of Mocapy++ with Biopython will provide a strong support for the field of protein structure prediction, design and simulation.
 +
;  Mentors
 +
[http://etal.myweb.uga.edu/ Eric Talevich]
 +
:  [http://wiki.binf.ku.dk/User:Thomas_Hamelryck Thomas Hamelryck]
  
; Approach : Rpy2 contains higher level interfaces to popular R libraries. For instance, the [http://rpy.sourceforge.net/rpy2/doc-2.1/html/graphics.html#package-ggplot2 ggplot2 interface] allows python users to access powerful plotting functionality in R with an intuitive API. Providing similar high level APIs for biological toolkits available in R would help expose these toolkits to a wider audience of Python programmers. A nice introduction to phylogenetic analysis in R is available from Rich Glor at the [http://bodegaphylo.wikispot.org/Phylogenetics_and_Comparative_Methods_in_R Bodega Bay Marine Lab wiki]. Some examples of R libraries for which integration would be welcomed are:
+
====  [http://biopython.org/wiki/GSOC2011_MocapyExt MocapyExt] ====
 +
; Student
 +
: Justinas V. Daugmaudis
 +
;  Rationale
 +
:  BioPython is a very popular library in Bioinformatics and Computational Biology. Mocapy++ is a machine learning toolkit for parameter learning and inference in dynamic Bayesian networks (DBNs), which encode probabilistic relationships among random variables in a domain. Mocapy++ is freely available under the GNU General Public Licence (GPL) from SourceForge. The library supports a wide spectrum of DBN architectures and probability distributions, including distributions from directional statistics. Notably, Kent distribution on the sphere and the bivariate von Mises distribution on the torus, which have proven to be useful in formulating probabilistic models of protein and RNA structure. Such a highly useful and powerful library, which has been used in such projects as TorusDBN, Basilisk, FB5HMM with great success, is the result of the long-term effort. The original Mocapy implementation dates back to 2004, and since then the library has been rewritten in C++. However, C++ is a statically typed and compiled programming language, which does not facilitate rapid prototyping. As a result, currently Mocapy++ has no provisions for dynamic loading of custom node types, and a mechanism to plug-in new node types that would not require to modify and recompile the library is of interest. Such a plug-in interface would assist rapid prototyping by allowing to quickly implement and test new probability distributions, which, in turn, could substantially reduce development time and effort; the user would be empowered to extend Mocapy++ without modifications and subsequent recompilations. Recognizing this need, the project (herein referred as MocapyEXT), with the aim to improve the current Mocapy++ node type extension mechanism, has been proposed by T. Hamelryck.
 +
;  Approach & Goals
 +
: The MocapyEXT project is largely an engineering effort to bring a transparent Python plug-in interface to Mocapy++, where built-in and dynamically loaded node types could be used in a uniform manner. Also, externally implemented and dynamically loaded nodes could be modified by a user and these changes will not necessitate the recompilation of the client program, nor the accompanying Mocapy++ library. This will facilitate rapid prototyping, ease the adaptation of currently existing code, and improve the software interoperability whilst introducing minimal changes to the existing Mocapy++ interface, thus facilitating a smooth acceptance of the changes introduced by MocapyEXT.
 +
;  Mentors
 +
[http://etal.myweb.uga.edu/ Eric Talevich]  
 +
:  [http://wiki.binf.ku.dk/User:Thomas_Hamelryck Thomas Hamelryck]
  
:* [http://ape.mpl.ird.fr/ ape (Analysis of Phylogenetics and Evolution)] -- an interactive library environment for phylogenetic and evolutionary analyses
+
=== 2010 ===
:* [http://pbil.univ-lyon1.fr/ADE-4/home.php?lang=eng ade4] -- Data Analysis functions to analyse Ecological and Environmental data in the framework of Euclidean Exploratory methods
+
====  [http://biopython.org/wiki/GSOC2010_Joao Improving Bio.PDB] ====
:* [http://cran.r-project.org/web/packages/geiger/index.html geiger] -- Running macroevolutionary simulation, and estimating parameters related to diversification from comparative phylogenetic data.
+
; Student
:* [http://picante.r-forge.r-project.org/ picante] -- R tools for integrating phylogenies and ecology
+
: [http://nmr.chem.uu.nl/~joaor João Rodrigues]
:* [http://mefa.r-forge.r-project.org/ mefa] -- multivariate data handling for ecological and biogeographical data
+
;  Rationale
 +
:  Biopython is a very popular library in Bioinformatics and Computational Biology. Its Bio.PDB module, originally developed by Thomas Hamelryck, is a simple yet powerful tool for structural biologists. Although it provides a reliable PDB parser feature and it allows several calculations (Neighbour Search, RMS) to be made on macromolecules, it still lacks a number of features that are part of a researcher's daily routine. Probing for disulphide bridges in a structure and adding polar hydrogen atoms accordingly are two examples that can be incorporated in Bio.PDB, given the module's clever structure and good overall organisation. Cosmetic operations such as chain removal and residue renaming – to account for the different existing nomenclatures – and renumbering would also be greatly appreciated by the community. Another aspect that can be improved for Bio.PDB is a smooth integration/interaction layer for heavy-weights in macromolecule simulation such as MODELLER, GROMACS, AutoDock, HADDOCK. It could be argued that the easiest solution would be to code hooks to these packages' functions and routines. However, projects such as the recently developed edPDB or the more complete Biskit library render, in my opinion, such interfacing efforts redundant. Instead, I believe it to be more advantageous to include these software' input/output formats in Biopython's SeqIO and AlignIO modules. This, together with the creation of interfaces for model validation/structure checking services/software would allow Biopython to be used as a pre- and post-simulation tool. Eventually, it would pave the way for its inclusion in pipelines and workflows for structure modelling, molecular dynamics, and docking simulations.
 +
;  Mentors
 +
: [http://etal.myweb.uga.edu/ Eric Talevich]  
 +
: [http://www.hutton.ac.uk/staff/peter-cock Peter Cock]
 +
:  Diana Jaunzeikare
  
; Challenges : The student would have the opportunity to learn an available R toolkit, and then code in Python and R to make this available via an intuitive API. This will involve digging into the R code examples to discover the most useful parts for analysis, and then projecting this into a library that is intuitive to Python coders. Beyond the coding and design aspects, the student should feel comfortable writing up use case documentation to support the API and encourage its adoption.
+
=== 2009 ===
  
; Involved toolkits or projects :
+
====  [http://biopython.org/wiki/PhyloXML PhyloXML] ====
 +
; Rationale
 +
:  PhyloXML is an XML format for phylogenetic trees, designed to allow storing information about the trees themselves (such as branch lengths and multiple support values) along with data such as taxonomic and genomic annotations. Connecting these pieces of evolutionary information in a standard format is key for comparative genomics.
 +
A Bioperl driver for phyloXML was created during the 2008 Summer of Code; this project aims to build a similar module for the popular Biopython package.
 +
;  Mentors
 +
:  [https://github.com/chapmanb Brad Chapman]
 +
: Christian Zmasek
  
:* [http://ape.mpl.ird.fr/ ape (Analysis of Phylogenetics and Evolution)]
+
====  [http://biopython.org/wiki/BioGeography Biogeographical Phylogenetics for BioPython] ====
:* [http://bitbucket.org/lgautier/rpy2/ Rpy2]
+
;  Rationale
:* [http://biopython.org/wiki/Main_Page Biopython]
+
:  I developed Bio.Geography, a new module for the bioinformatics programming toolkit Biopython. Bio.Geography expands upon Biopython's traditional capabilities for accessing gene and protein sequences from online databases by allowing automated searching, downloading, and parsing of geographic location records from GBIF, the authoritative aggregator of specimen information from natural history collections worldwide. This will enable analyses of evolutionary biogeography that require the areas inhabited by the species at the tips of the phylogeny, particularly for large-scale analyses where it is necessary to process thousands of specimen occurrence records. The module will also facilitate applications such as species mapping, niche modeling, error-checking of museum records, and monitoring range changes.
 +
;  Mentors
 +
: [https://github.com/chapmanb Brad Chapman]
 +
: Stephen Smith
 +
: David Kidd
  
; Degree of difficulty and needed skills : Moderate. The project requires familiarity with coding in Python and R, and knowledge of phylogeny or biogeography. The student has plenty of flexibility to define the project based on their biological interests (e.g. [http://www.warwick.ac.uk/go/peter_cock/python/heatmap/ microarrays and heatmaps]); there is also the possibility to venture far into data visualization once access to analysis methods is made. [http://kiwi.cs.dal.ca/GenGIS/Main_Page GenGIS] and can give ideas about what is possible.
 
  
; Mentors : [http://dk.linkedin.com/pub/laurent-gautier/8/81/869 Laurent Gautier], [http://bcbio.wordpress.com Brad Chapman], [http://www.scri.ac.uk/staff/petercock Peter Cock]
+
== Future project ideas ==
  
=== Synergy from Biopython and Mocapy++ ===
+
The BioPython proposals for future Google Summers of Code are to be published here once discussed. We encourage potential students and mentors to join the [http://biopython.org/wiki/Mailing_lists BioPython mailing lists] and actively participate in these discussions, either by submitting their own ideas or contributing to improving existing ones.
  
; Rationale : [http://sourceforge.net/projects/mocapy/ Mocapy++] is a machine learning toolkit for training and using [http://en.wikipedia.org/wiki/Bayesian_network Bayesian networks]. Mocapy++ supports the use of [http://en.wikipedia.org/wiki/Directional_statistics directional statistics]; the statistics of angles, orientations and directions. This unique feature of Mocapy++ makes the toolkit especially suited for the formulation of probabilistic models of biomolecular structure. The toolkit has already been used to develop (published and peer reviewed) models of [http://www.pnas.org/content/105/26/8932.abstract?etoc protein] and [http://www.ploscompbiol.org/article/info%3Adoi%2F10.1371%2Fjournal.pcbi.1000406 RNA] structure in atomic detail. Mocapy++ is implemented in C++, and does not provide any Python Bindings. The goal of this proposal is to develop an easy-to-use Python interface to Mocapy++, and to integrate this interface with the Biopython project. Through its [http://biopython.org/DIST/docs/cookbook/biopdb_faq.pdf Bio.PDB] module (initially implemented by the mentor of this proposal, [http://www.binf.ku.dk/research/structural_bioinformatics/ T. Hamelryck]), Biopython provides excellent functionality for data mining of biomolecular structure databases. Integrating Mocapy++ and Biopython would create strong synergy, as it would become quite easy to extract data from the databases, and subsequently use this data to formulate probabilistic models. As such, it would provide a strong impulse to the field of protein structure prediction, design and simulation. Possible applications beyond bioinformatics are obvious, and include probabilistic models of human or animal movement, or any other application that involves directional data.
+
=== Indexing & Lazy-loading Sequence Parsers ===
 +
; Rationale
 +
: [[SeqIO|Bio.SeqIO]]'s indexing offers parsing on demand access to any sequence in a large file (or collection of files on disk) as a [[SeqRecord]] object. This works well when you have many small to medium sized sequences/genomes. However, this is not ideal for large genomes or chromosomes where only a sub-region may be needed. A lazy-loading parser would delay reading the record until requested. For example, if region ''record[3000:4000]'' is requested, then only those 1000 bases need to be loaded from disk into memory, plus any features in that region. This is how Biopython's [[BioSQL]] interface works. Tools like tabix and samtools have demonstrated efficient co-ordinate indexing which could be useful here.
 +
: Aside from being used via an index for random access, lazy-loading parsers could be used when iterating over a file as well. This can ''potentially'' offer speed ups for tasks where only a fraction of the data is used. For example, if calculating the GC content of a collection of genomes from GenBank, using Bio.SeqIO.parse(...) would currently needlessly load and parse all the annotation and features. A lazy-parser would only parse the sequence information.
 +
;  Approach & Goals
 +
: Useful features include:
 +
* Internal indexing of multiple file formats, including FASTA and richly annotated sequence formats like GenBank/EMBL and GTF/GFF/GFF3.
 +
* Full compatibility with existing SeqIO parsers which load everything into memory as a `SeqRecord` object.
 +
;  Difficulty and needed skills
 +
: Hard. Familiarity with the Biopython's existing sequence parsing essential. Understanding of indexing large files will be vital.
 +
;  Possible Mentors
 +
: [https://github.com/peterjc/ Peter Cock], others welcome
  
; Approach : Ideally, the student would first gain some understanding of the theoretical background of the algorithms that are used in Mocapy++, which involves parameter learning of Bayesian networks using [http://en.wikipedia.org/wiki/Expectation-maximization_algorithm Stochastic Expectation Maximization (S-EM)]. Next, the student would study some of the use cases of the toolkit, making use of some of the published articles that involve Mocapy++. After becoming familiar with the internals of Mocapy++, Python bindings will then be implemented using the Boost C++ library. Based on the use cases, the student would finally implement some example applications that involve data mining of biomolecular structure using Biopython, the subsequent formulation of probabilistic models using Python-Mocapy++, and its application to some biologically relevant problem. Schematically, the following steps are involved for the student: 
 
  
:* Theoretical study of S-EM
+
<!--
:* Study of Mocapy++ use cases
+
=== XXXX ===
:* Study of Mocapy++ internals and code
+
====  Mock Proposal ====
:* Implementing Python bindings using Boost
+
;  Rationale
:* Example applications, involving Bio.PDB data mining
+
: aaa
 
+
; Approach & Goals
; Challenges : The project is highly interdisciplinary, and ideally requires skills in programming (C++, Python, wrapping C++ libraries in Python, Boost), machine learning, knowledge of biomolecular structure and statistics. The project could be extended (for example, by implementing additional functionality in Mocapy++) or limited (for example by limiting the time spent on understanding the theory behind Mocapy++).
+
: zzz
 
+
; Difficulty and needed skills
; Involved toolkits or projects :
+
: yyy
 
+
; Possible Mentors
:* [http://biopython.org/wiki/Main_Page Biopython]
+
: xxx
:* [http://sourceforge.net/projects/mocapy/ Mocapy++]
+
-->
 
+
; Degree of difficulty and needed skills : Hard. The student needs to be fluent in C++, Python and the [http://www.boost.org C++ Boost library]. Experience with machine learning, Bayesian statistics and biomolecular structure would be clear advantages.
+
 
+
; Mentors : [http://www.binf.ku.dk/research/structural_bioinformatics/ Thomas Hamelryck]
+

Revision as of 03:55, 14 June 2013

Contents

Introduction

For the past several years, Biopython-related GSoC projects have been successfully run under the mentorship of the Open Bioinformatics Foundation (OBF) and the National Evolutionary Synthesis Center (NESCent).

Please read those organizations' GSoC pages and the main Google Summer of Code page for more details about the program.

Mentor List

Usually, each BioPython proposal has one or more mentors assigned to it. Nevertheless, we encourage potential students/mentors to contact the mailing lists with their own ideas for proposals. There is therefore not a set list of 'available' mentors, since it highly depends on which projects are proposed every year.

Past mentors include:


Current projects

This year, NESCent was selected as a mentoring organization for Google Summer of Code. See also NESCent's GSoC page.

Codon alignment and analysis

Student
Zheng Ruan
Rationale
A codon alignment is an alignment of nucleotide sequences in which the trinucleotides correspond directly to amino acids in the translated protein product. This carries important information which can be used for several analysis, notably analysis of selection pressures. This project extends Biopython to support this data type and these analyses.
Approach & Goals
Useful features include:
  • Conversion of a set of unaligned nucleic acid sequences and a corresponding protein sequence alignment to a codon alignment.
  • Calculation of selection pressure from the ratio of nonsynonomous to synonomous site replacements, and related functions.
  • Model selection.
  • Possible extension of AlignIO and the MultipleSeqAlignment class to take full advantage of codon alignments, including validation (testing for frame shifts, etc.)
Difficulty and needed skills
Medium, depending on ambition. Familiarity with the Biopython's existing alignment classes and functions, or equivalents in BioPerl, BioJava or BioRuby (e.g.), will be helpful. Understanding of the practical uses of codon alignments, or at least a basic understanding of molecular biology, is important. Some basic math is involved, essentially reading a few equations and converting them to code. One useful book to have on hand is Computational Molecular Evolution by Ziheng Yang.
Mentors
Eric Talevich, Peter Cock

Bio.Phylo: filling in the gaps

Student
Yanbo Ye
Rationale
While the Phylo module in Biopython supports I/O and basic tree operations, there are some important components that remain to be implemented to better support phylogenetic workflows.
Approach & Goals
This "idea" is intentially left open-ended -- some potentially useful features are:
  • A "Phylo.consensus" module with functions for the consensus of multiple trees. E.g.: Strict consensus, as Bio.Nexus already implements; perhaps other methods like Adams consensus.
  • A function to calculate bootstrap support for branches given a target or "master" tree and a series of bootstrap replicate trees (usually read incrementally from a Newick file).
  • Functions for comparison of two or more trees.
  • Simple algorithms for tree inference, i.e. neighbor-joining and parsimony tree estimation. For small alignments (and perhaps medium-sized ones with PyPy), it would be nice to run these without an external program, e.g. to construct a guide tree for another algorithm or quickly view a phylogenetic clustering of sequences.
  • Tree visualizations: A proper draw_unrooted function to perform radial layout, with an optional "iterations" argument to use Felsenstein's Equal Daylight algorithm. Circular radial diagrams are also popular these days. Any new function should support the same arguments as the existing Phylo.draw.
Difficulty and needed skills
Medium-to-advanced programming skill in Python -- it's important for these implementations to be reasonably efficient, though we don't aim to compete with the fastest stand-alone implementations of these algorithms. Knowledge of phylogenetic methods is critical; for reference, you might like to have a copy of Joe Felsenstein's Inferring Phylogenies. Tree visualizations are done with matplotlib.
Mentors
Eric Talevich, others welcome


Past Proposals

2012

SearchIO

Rationale
Biopython has general APIs for parsing and writing assorted sequence file formats (SeqIO), multiple sequence alignments (AlignIO), phylogenetic trees (Phylo) and motifs (Bio.Motif). An obvious omission is something equivalent to BioPerl's SearchIO. The goal of this proposal is to develop an easy-to-use Python interface in the same style as SeqIO, AlignIO, etc but for pairwise search results. This would aim to cover EMBOSS muscle & water, BLAST XML, BLAST tabular, HMMER, Bill Pearson's FASTA alignments, and so on.
Approach
Much of the low level parsing code to handle these file formats already exists in Biopython, and much as the SeqIO and AlignIO modules are linked and share code, similar links apply to the proposed SearchIO module when using pairwise alignment file formats. However, SearchIO will also support pairwise search results where the pairwise sequence alignment itself is not available (e.g. the default BLAST tabular output). A crucial aspect of this work will be to design a pairwise-search-result object heirachy that reflects this, probably with a subclass inheriting from both the pairwise-search-result and the existing MultipleSequenceAlignment object. Beyond the initial challenge of an iterator based parsing and writing framework, random access akin to the Bio.SeqIO.index and index_db functionality would be most desirable for working with large datasets.
Challenges
The project will cover a range of important file formats from major Bioinformatics tools, thus will require familiarity with running these tools, and understanding their output and its meaning. Inter-converting file formats is part of this.
Difficulty and needed skills
Medium/Hard depending on how many objectives are attempted. The student needs to be fluent in Python and have knowledge of the BioPython codebase. Experience with all of the command line tools listed would be clear advantages, as would first hand experience using BioPerl's SearchIO. You will also need to know or learn the git version control system.
Mentors
Peter Cock

Representation and manipulation of genomic variants

Rationale
Computational analysis of genomic variation requires the ability to reliably communicate and manipulate variants. The goal of this project is to provide facilities within BioPython to represent sequence variation objects, convert them to and from common human and file representations, and provide common manipulations on them.
Approach & Goals
  • Object representation
    • identify variation types to be represented (SNV, CNV, repeats, inversions, etc)
    • develop internal machine representation for variation types
    • ensure coverage of essential standards, including HGVS, GFF, VCF
  • External representations
    • write parser and generators between objects and external string and file formats
  • Manipulations
    • canonicalize variations with more than one valid representation (e.g., ins versus dup and left shifting repeats).
    • develop coordinate mapping between genomic, cDNA, and protein sequences (HGVS)
  • Other
    • release code to appropriate community efforts and write short manuscript
    • implement web service for HGVS conversion
Difficulty and needed skills
Easy-to-Medium depending on how many objectives are attempted. The student will need have skills in most or all of: basic molecular biology (genomes, transcripts, proteins), genomic variation, Python, BioPython, Perl, BioPerl, NCBI Eutilities and/or Ensembl API. Experience with computer grammars is highly desirable. You will also need to know or learn the git version control system.
Mentors
Reece Hart
Brad Chapman
James Casbon

2011

Biomolecular Interface Analysis

Student
Mikael Trellet
Rationale
Analysis of protein-protein complexes interfaces at a residue level yields significant information on the overall binding process. Such information can be broadly used for example in binding affinity studies, interface design, and enzymology. To tap into it, there is a need for tools that systematically and automatically analyze protein structures, or that provide means to this end. Protorop (http://www.bioinformatics.sussex.ac.uk/protorp/) is an example of such a tool and the elevated number of citations the server has had since its publication acknowledge its importance. However, being a webserver, Protorop is not suited for large-scale analysis and it leaves the community dependent on its maintainers to keep the service available. On the other hand, Biopython’s structural biology module, Bio.PDB, provides the ideal parsing machinery and programmatic structures for the development of an offline, open-source library for interface analysis. Such a library could be easily used in large-scale analysis of protein-protein interfaces, for example in the CAPRI experiment evaluation or in benchmark statistics. It would be also reasonable, if time permits, to extend this module to deal with protein-DNA or protein-RNA complexes, as Biopython supports nucleic acids already.
Approach & Goals
  • Add the new module backbone in current Bio.PDB code base
    • Evaluate possible code reuse and call it into the new module
    • Try simple calculations to be sure that there is stability between the different modules (parsing for example) and functions
  • Define a stable benchmark
    • Select few PDB files among interface size and proteins size would be different
  • Extend IUPAC.Data module with residue information
    • Deduce residues weight from Atom instead of direct dictionary storage
    • Polar/charge character (dictionary or influenced by pH)
    • Hydrophobicity scale(s)
  • Implement Extended Residue class as a subclass of Residue
  • Implement Interface object and InterfaceAnalysis module
  • Develop functions for interface analysis
    • Calculation of interface polar character statistics (% of polar residues, apolar, etc)
    • Calculation of BSA calling MSMS or HSA
    • Calculation of SS element statistics in the interface through DSSP
    • Unit tests and use of results as input for further calculations by other tools and scripts
  • Develop functions for Interface comparison
  • Code organization and final testing
Difficulty and needed skills
Easy/Medium. Working knowledge of the Bio.PDB module of BioPython. Knowledge of structural biology in general and associated file formats (PDB).
Mentors
João Rodrigues
Eric Talevich

A Python bridge for Mocapy++

Student
Michele Silva
Rationale
Discovering the structure of biomolecules is one of the biggest problems in biology. Given an amino acid or base sequence, what is the three dimensional structure? One approach to biomolecular structure prediction is the construction of probabilistic models. A Bayesian network is a probabilistic model composed of a set of variables and their joint probability distribution, represented as a directed acyclic graph. A dynamic Bayesian network is a Bayesian network that represents sequences of variables. These sequences can be time-series or sequences of symbols, such as protein sequences. Directional statistics is concerned mainly with observations which are unit vectors in the plane or in three-dimensional space. The sample space is typically a circle or a sphere. There must be special directional methods which take into account the structure of the sample spaces. The union of graphical models and directional statistics allows the development of probabilistic models of biomolecular structures. Through the use of dynamic Bayesian networks with directional output it becomes possible to construct a joint probability distribution over sequence and structure. Biomolecular structures can be represented in a geometrically natural, continuous space. Mocapy++ is an open source toolkit for inference and learning using dynamic Bayesian networks that provides support for directional statistics. Mocapy++ is excellent for constructing probabilistic models of biomolecular structures; it has been used to develop models of protein and RNA structure in atomic detail. Mocapy++ is used in several high-impact publications, and will form the core of the molecular modeling package Phaistos, which will be released soon. The goal of this project is to develop a highly useful Python interface to Mocapy++, and to integrate that interface with the Biopython project. Through the Bio.PDB module, Biopython provides excellent functionality for data mining biomolecular structure databases. Integrating Mocapy++ and Biopython will allow training a probabilistic model using data extracted from a database. Integrating Mocapy++ with Biopython will create a powerful toolkit for researchers to quickly implement and test new ideas, try a variety of approaches and refine their methods. It will provide strong support for the field of biomolecular structure prediction, design, and simulation.
Approach & Goals
Mocapy++ is a machine learning toolkit for training and using Bayesian networks. It has been used to develop probabilistic models of biomolecular structures. The goal of this project is to develop a Python interface to Mocapy++ and integrate it with Biopython. This will allow the training of a probabilistic model using data extracted from a database. The integration of Mocapy++ with Biopython will provide a strong support for the field of protein structure prediction, design and simulation.
Mentors
Eric Talevich
Thomas Hamelryck

MocapyExt

Student
Justinas V. Daugmaudis
Rationale
BioPython is a very popular library in Bioinformatics and Computational Biology. Mocapy++ is a machine learning toolkit for parameter learning and inference in dynamic Bayesian networks (DBNs), which encode probabilistic relationships among random variables in a domain. Mocapy++ is freely available under the GNU General Public Licence (GPL) from SourceForge. The library supports a wide spectrum of DBN architectures and probability distributions, including distributions from directional statistics. Notably, Kent distribution on the sphere and the bivariate von Mises distribution on the torus, which have proven to be useful in formulating probabilistic models of protein and RNA structure. Such a highly useful and powerful library, which has been used in such projects as TorusDBN, Basilisk, FB5HMM with great success, is the result of the long-term effort. The original Mocapy implementation dates back to 2004, and since then the library has been rewritten in C++. However, C++ is a statically typed and compiled programming language, which does not facilitate rapid prototyping. As a result, currently Mocapy++ has no provisions for dynamic loading of custom node types, and a mechanism to plug-in new node types that would not require to modify and recompile the library is of interest. Such a plug-in interface would assist rapid prototyping by allowing to quickly implement and test new probability distributions, which, in turn, could substantially reduce development time and effort; the user would be empowered to extend Mocapy++ without modifications and subsequent recompilations. Recognizing this need, the project (herein referred as MocapyEXT), with the aim to improve the current Mocapy++ node type extension mechanism, has been proposed by T. Hamelryck.
Approach & Goals
The MocapyEXT project is largely an engineering effort to bring a transparent Python plug-in interface to Mocapy++, where built-in and dynamically loaded node types could be used in a uniform manner. Also, externally implemented and dynamically loaded nodes could be modified by a user and these changes will not necessitate the recompilation of the client program, nor the accompanying Mocapy++ library. This will facilitate rapid prototyping, ease the adaptation of currently existing code, and improve the software interoperability whilst introducing minimal changes to the existing Mocapy++ interface, thus facilitating a smooth acceptance of the changes introduced by MocapyEXT.
Mentors
Eric Talevich
Thomas Hamelryck

2010

Improving Bio.PDB

Student
João Rodrigues
Rationale
Biopython is a very popular library in Bioinformatics and Computational Biology. Its Bio.PDB module, originally developed by Thomas Hamelryck, is a simple yet powerful tool for structural biologists. Although it provides a reliable PDB parser feature and it allows several calculations (Neighbour Search, RMS) to be made on macromolecules, it still lacks a number of features that are part of a researcher's daily routine. Probing for disulphide bridges in a structure and adding polar hydrogen atoms accordingly are two examples that can be incorporated in Bio.PDB, given the module's clever structure and good overall organisation. Cosmetic operations such as chain removal and residue renaming – to account for the different existing nomenclatures – and renumbering would also be greatly appreciated by the community. Another aspect that can be improved for Bio.PDB is a smooth integration/interaction layer for heavy-weights in macromolecule simulation such as MODELLER, GROMACS, AutoDock, HADDOCK. It could be argued that the easiest solution would be to code hooks to these packages' functions and routines. However, projects such as the recently developed edPDB or the more complete Biskit library render, in my opinion, such interfacing efforts redundant. Instead, I believe it to be more advantageous to include these software' input/output formats in Biopython's SeqIO and AlignIO modules. This, together with the creation of interfaces for model validation/structure checking services/software would allow Biopython to be used as a pre- and post-simulation tool. Eventually, it would pave the way for its inclusion in pipelines and workflows for structure modelling, molecular dynamics, and docking simulations.
Mentors
Eric Talevich
Peter Cock
Diana Jaunzeikare

2009

PhyloXML

Rationale
PhyloXML is an XML format for phylogenetic trees, designed to allow storing information about the trees themselves (such as branch lengths and multiple support values) along with data such as taxonomic and genomic annotations. Connecting these pieces of evolutionary information in a standard format is key for comparative genomics.

A Bioperl driver for phyloXML was created during the 2008 Summer of Code; this project aims to build a similar module for the popular Biopython package.

Mentors
Brad Chapman
Christian Zmasek

Biogeographical Phylogenetics for BioPython

Rationale
I developed Bio.Geography, a new module for the bioinformatics programming toolkit Biopython. Bio.Geography expands upon Biopython's traditional capabilities for accessing gene and protein sequences from online databases by allowing automated searching, downloading, and parsing of geographic location records from GBIF, the authoritative aggregator of specimen information from natural history collections worldwide. This will enable analyses of evolutionary biogeography that require the areas inhabited by the species at the tips of the phylogeny, particularly for large-scale analyses where it is necessary to process thousands of specimen occurrence records. The module will also facilitate applications such as species mapping, niche modeling, error-checking of museum records, and monitoring range changes.
Mentors
Brad Chapman
Stephen Smith
David Kidd


Future project ideas

The BioPython proposals for future Google Summers of Code are to be published here once discussed. We encourage potential students and mentors to join the BioPython mailing lists and actively participate in these discussions, either by submitting their own ideas or contributing to improving existing ones.

Indexing & Lazy-loading Sequence Parsers

Rationale
Bio.SeqIO's indexing offers parsing on demand access to any sequence in a large file (or collection of files on disk) as a SeqRecord object. This works well when you have many small to medium sized sequences/genomes. However, this is not ideal for large genomes or chromosomes where only a sub-region may be needed. A lazy-loading parser would delay reading the record until requested. For example, if region record[3000:4000] is requested, then only those 1000 bases need to be loaded from disk into memory, plus any features in that region. This is how Biopython's BioSQL interface works. Tools like tabix and samtools have demonstrated efficient co-ordinate indexing which could be useful here.
Aside from being used via an index for random access, lazy-loading parsers could be used when iterating over a file as well. This can potentially offer speed ups for tasks where only a fraction of the data is used. For example, if calculating the GC content of a collection of genomes from GenBank, using Bio.SeqIO.parse(...) would currently needlessly load and parse all the annotation and features. A lazy-parser would only parse the sequence information.
Approach & Goals
Useful features include:
  • Internal indexing of multiple file formats, including FASTA and richly annotated sequence formats like GenBank/EMBL and GTF/GFF/GFF3.
  • Full compatibility with existing SeqIO parsers which load everything into memory as a `SeqRecord` object.
Difficulty and needed skills
Hard. Familiarity with the Biopython's existing sequence parsing essential. Understanding of indexing large files will be vital.
Possible Mentors
Peter Cock, others welcome


Personal tools
Namespaces
Variants
Actions
Navigation
Toolbox