Current Features

  • Easy to install with pip
  • Easy to use, with most common functions at top package level
  • Good documentation and tutorials
  • Extensible API, allowing for bioinformaticians to develop their own algorithms and own their data, instead of using a complicated plugin in Cytoscape with multiple dependencies and little ability for export or customization
  • Caching system for downloaded namespace and annotations
  • Fully BEL 1.0 and BEL 2.0 support
  • Node and edge filtering API
  • Usage of OWL documents as namespaces


  • Speed is still an issue, because documents above 100K lines still take a couple minutes to run. This issue is exacerbated by (optionally) logging output to the console, which can make it more than 3x or 4x as slow.
  • The default namespaces from OpenBEL do not follow a standard file format. They are similar to INI config files, but do not use consistent delimiters. Also, many of the namespaces don’t respect that the delimiter should not be used in the namespace names. There are also lots of names with strange characters, which may have been caused by copying from a data source that had specfic escape characters without proper care.
  • Testing was very difficult because the example documents on the OpenBEL website had many semantic errors, such as using names and annotation values that were not defined within their respective namespace and annotation definition files. They also contained syntax errors like naked names, which are not only syntatically incorrect, but lead to bad science; and improper usage of activities, like illegally nesting an activity within a composite statement.



  • Parallelization of BEL tokenization

Knowledge Assembly

  • Integration of prior knowledge, such orthology between genes/proteins and resolution of protein complex names
  • Integration of heirarchical information, insertion of missing complex members and protein family heirarchies
  • Leverage of ontologies in the OWL format for disease-specific entities and annotations

Post-Processing Tools

  • Implementation of Reverse Causal Reasoning and Network Perturbation Amplitude
  • Distributed BEL document infrastructure
  • Semantic Diff tool

Data Interchange

Another format, Biological Pathways Expression Language (BioPax), was developed to integrate pathway information from many disparate databases. Like BEL, it can describe metabolic pathways and molecular interactions, but it excels in signaling pathways, gene regulatory networks, and genetic interactions. Systems Biology Markup Language (SBML) is the third common format that provides a more general framework for building quantitative and temporal models.

Data locked away in other formats such as BioPax and SBML cannot be accessed by PyBEL currently. Development of knowledge assemblers, like INDRA, provide support for import of many formats. PyBEL will enable the import of BEL documents much more quickly, and ultimately enable the export of SMBL. In the future, it would also be useful to develop additional interchange tools for BioPax to BEL, but we recognize that this is a large task that will be limited by the expressibility of each language and the difficult development of a two-way mapping.