gsoc13


Ideas for GSoC 2013 projects in Crowdsourcing Biology

  • Idea 0. Create your own!
  • Idea 1.  Gene Wiki. Convert Gene Wiki bot to write to WikiData
  • Idea 2. Games. Develop a new version of the game Dizeez as a Facebook application
  • Idea 3. BioGPS. New gene query interface (backed by MyGene.info v2 API)
  • Idea 4. BioGPS. Advanced gene annotation tools: ID mapping and user table annotation.
  • Idea 5. BioGPS. Gene report layout “sharing by link” button
  • Idea 6. Games. Add interactive decision tree builder for the The Cure
  • Idea 7. BioGPS. Add plugin to display content from Semantic BioGPS.

(A pre-approval, fully open ideas google document is available, but the ideas listed below reflect our current priority)

Idea 0. Create your own!

Feel free to propose your own idea.  As long as it fits within the general space of ‘crowdsourcing biology’ we would be thrilled to consider it.  Please be detailed about what the specific goal would be, why it is important and how you plan to achieve it.

Idea 1. Convert Gene Wiki bot to write to WikiData

Background. Code developed by our group helps to maintain more than 10,000 articles about human genes on Wikipedia.  All of the page elements highlighted in orange below are generated by a program that gathers information from trusted public databases (accessed via MyGene.info) and inserts it into the right structure at Wikipedia.  Collectively the articles are viewed more than 50 million times a year.

Wikidata is a brand new development at Wikipedia.  Its goal is to create a free database that anyone can edit.  The aim is to collect all of the “structured data”, the elements of infoboxes for example, into a central, queriable database.  Articles on Wikipedia can query this database to access and then render the content much like other web applications.  In addition, external web applications can dynamically query this content to enhance their own content.  The goal of this project will be to migrate the data that we currently store in template pages in Wikipedia into the wikidata database.

For example, we now get data from mygene.info and add it to templates like this one for the gene Reelin .  We would like to move that data instead into the wikidata page associated with Reelin.  This will allow us to query the database, others to edit the database, etc. in a much cleaner fashion than we can accomplish working inside of the Wikipedia article/template text directly.  This will also form the basis for adding new structured content about genes into Wikipedia such as explicit connections between genes and disease.

Specifics. Convert the pygenewikibot to deposit the data used to populate the gene articles into WikiData. (see issue tracker).  Add visualization elements to one or both of the gene wiki (Wikipedia) and GeneWiki+ that display the information from wikidata.

You should consider this project if: you know Python, you are excited about working to make Wikipedia better, and you want your work to be used by millions of people every day.

Possible mentors: Ben, Sal, Max, Andrew

References: Molecular Biology task force for wikidata, a repo for wikidata bot code, gene wiki portal

Idea 2. Develop a new version of the game Dizeez as a Facebook application

Background. Dizeez is a simple quiz game about connections between genes and diseases.  Players learn about gene-disease connections through game play.  In addition, the system learns new connections suggested by many different players – thus providing a mechanisms to crowdsource the creation of a new gene-disease database.  Building on some early success with a simple Dizeez webapp, this project will bring this game into the Facebook context – hopefully garnering many more players, creating a more enjoyable player experience, and producing better data.

Suggested implementation:  Facebook Canvas application

  1. Built as a single page javascript application
    1. Marionette (w/ Backbone) using RequireJS
  2. API to sync game play results and get back new gene & disease options.
  3. FB Integration
    1. Score API to challenge players in your social graph
    2. Achievements API to promote level and game advancements
    3. Notifications to get others to join you
  4. Dynamic disease choice answer compositions

Mentors: Max

Idea 3. BioGPS. New gene query interface (backed by MyGene.info v2 API)

MyGene.info v2 gene query API expands both data and query capabilities over the current APIs. We designed a new and clean query interface to utilize these new APIs. Here are a few highlights (with mock-ups provided when necessary):

  1. By default, it supports a single line query input with a rich query syntax (as provided by MyGene.info API)  (mockup)
  2. A gene query result page to show the list of hits returned from query API. (mockup)
  3. Support Advanced query mode for multi-line queries (e.g. using a “area” element instead of “input”)
    • mode 1: auto-switch to multi-line query mode when users paste into a multi-line text into default input box.
    • mode 2: switch to multi-line query mode by a link or button. if mode 1 is hard to implement, this will be the fallback option.
  4. Besides default queries on genes, BioGPS also support queries on other entities, like plugins, datasets, genelists. It will be nice to implement a “Google-Drive” style query options/filters (see “images” filter below)

Refer to this post for setting up a BioGPS dev environment. Also note that, if preferred, this project can be developed as a standalone application as well (without BioGPS dev environment). The integration with BioGPS can be done in later stage.

Mentors: Chunlei, Andrew

http://biogps.org

http://mygene.info

Additional information: http://genomebiology.com/2009/10/11/R130 http://nar.oxfordjournals.org/content/41/D1/D561.short

 

Idea 4. BioGPS. Advanced gene annotation tools: ID mapping and User table annotation.

Idea 3 above implements the general query interface for BioGPS. Besides that, we also want to implement two commonly used gene annotation features:

  1. ID mapping Users can copy/paste or load (from a file) a list of gene identifiers,  and then specify one or multiple identifier types (Entrez GeneID, RefSeq ID, Uniprot ID, etc), the fields they want to output (the default will be “symbol”, “name”, can be any fields supported by MyGene.info API). The output will be a table of matching genes with user-specified output fields.
  2. User table annotation This is tightly related with “ID mapping”. The difference is that users will load a table from a file (tsv, csv, xlsx, etc), then specify which column is the “identifer” column they want to query for the gene annotations. They will then specify the identifier types and output fields as well. The output fields will be inserted into users’ table as additional columns. A “download” link will be provided for users to download the entire annotated table.

Mentors: Chunlei, Andrew

http://biogps.org

http://mygene.info

Additional information: http://genomebiology.com/2009/10/11/R130 http://nar.oxfordjournals.org/content/41/D1/D561.short

 

Idea 5. BioGPS. Gene report layout “sharing by link” button

BioGPS’s gene report page (like this one) is the core landing page for our users, which displays variety of gene-centric resources users picked from our plugin library. We use the concept of “layout” (or “gene report layout”)  to represent the view of users’ customized gene report page, which essentially contains the information of plugins identities, window poistions/sizes, etc. This is an example layout. We would like to implement a layout-sharing feature described as the following:

  1. It works similar to google map�����s sharing button below, we will put a “sharing by link” button in gene report page.
  2. When the button was clicked, a “long” url is generated, which contains all information to re-produce user�������������������s current gene report view.
  3. Allow users to share the link via common social network (twitter, facebook, etc.)
  4. Allow to generate a persistent “short-url” version of the long one (http://biogps.org/view/qMjXHuaG). Another idea is to delegate this feature to external “url shortener” service.

Refer to this post for setting up a BioGPS dev environment.

Mentors: Chunlei, Andrew

Additional information: http://genomebiology.com/2009/10/11/R130 http://nar.oxfordjournals.org/content/41/D1/D561.short

Idea 6. Games. Add interactive decision tree builder to The Cure

Background.  The Cure is a game related to machine learning and classification in biological domains (especially breast cancer).  In the current implementation, a decision tree is automatically constructed, tested and rendered by the system based only on the player picking which variables (genes) to use to build it. Idea. Create a new game interface that lets the players interactively construct the decision trees based on any number of genes of their choosing. This would basically be an ‘expert mode’ that gave players full control. The trees would be evaluated (server-side) based on how well they classified the training set, how big they were (smaller is better), and how biologically meaningful the group of genes in the tree is.

See repo.

curetree

Suggested implementation.  D3 javascript.

Mentors: Max, Ben


Idea 7. BioGPS. Add plugin to display content from Semantic BioGPS.

Background: The broad aim of this project is to enable information extraction from all of the websites (plugins) accessible via biogps.org .  Specifically, we are attempting to gather gene annotations.  For example triples like: “Gene X plays a role in Alzheimer���s Disease”.  To extract this knowledge, we have a Web interface that allows users to visually construct HTML parsers specific to each BioGPS plugin (that end up being Xpath statements).  There is also code that runs these parsers over large volumes of gene-centric pages and stores the extracted content in a MongoDB database.  (See https://bitbucket.org/sulab/semantic-biogps)

Idea: For this year’s GSoC, we propose the development of a Semantic BioGPS display plugin.  This would query the data extracted from all of the annotated plugins and display it in the context of BioGPS.  The data to be displayed would be in a triple structure with the gene as the subject and could have a large variety of properties (predicates) and object types. While this interface is the end goal, the project would also need to improve on the current annotation authoring pipeline to ensure that it is robust enough to support a public interface.  This will likely involve refinements to the annotation authoring interface and regularly scheduled crawls of the biogps plugin library. ��Because of the nature of the task (parsing structured data from semi-structured and transient HTML) robust error-handling is vital to consider at every step.

Possible Mentors: Chunlei, Andrew, Ben, Max

Possible interface dev tools: roll your own html/javascript, Exhibit, D3, Cytoscapeweb

A demo site for semantic BioGPS is live here: http://54.244.135.254/

Additional information: http://genomebiology.com/2009/10/11/R130 http://nar.oxfordjournals.org/content/41/D1/D561.short