Kozo Nishida - Automating biological network visualization with Jupyter Notebook | JupyterCon 2020

Key Takeaways

hi i'm kozo these slides and speaker notes were created by me but amazon poly matthew will automatically read it in this talk i will show that automation of biological network visualization can be achieved by leveraging jupiter notebook this is the outline of this talk first i give you an overview of biological network visualization next i will explain why it would be nice to automate the visualization with jupiter notebook next i will show you a demo of automation using jupiter notebook finally i will explain what is lacking in this jupyter notebook application and tell you the future prospects of it this table is an example data for a network nodes a row represents a biomolecule node and a column represents some attribute information of a biomolecule in addition to the node data there is also edge data that represents their relationships biological network visualization is to create a network diagram like the figure on the right from the data on the left in the network figure visual effects are given according to the cell values in the table and some biological knowledge can be obtained from the network visualization site escape is the most commonly used software for biological network visualization site escape is a desktop application that imports and visualizes the data described in the previous slide site escape is written in java and runs on windows mac os and linux site escape is a popular software but site escape itself is not enough for the real world workflow in order to create site escape input data it is necessary to acquire data from multiple data sources and perform pre-processing outside site escape since site escape is a desktop application we need to perform some manual operation with a graphical user interface we may also do some analysis after loading the data into site escape site escape has some analytic capabilities but that may not be enough after manually manipulating the functions that site escape does not have we can finally obtain publishing quality images in fact site escape has a mechanism like a google chrome extension to supplement the features that cite escape does not have it's called the site escape app the scope of its functionality is not necessarily in this red frame but the app makes it easy to perform some workflow with features that cite escape does not have the site escape app like side escape must be implemented in java however there are still issues with the site escape app first of all it's difficult to implement the site escape app must be implemented in java but java is not the primary language for bio researchers the primary language for bio researchers is python or r second there are a few cytoscape apps that will run the entire workflow for this reason some manual work is required after all this is a problem in terms of research reproducibility so i use jupiter notebook and python or r to promote site escape workflow automation site escape is a desktop application but it also becomes a rest server as soon as it starts by sending rest requests we can programmatically perform site escape operations that would normally be done in the graphical user interface by utilizing this mechanism the workflow can be executed automatically in python or r python and r wrappers for the rest requests are released as packages called pi 4 side escape and rsi 3 respectively i'm running pi 4 side escape or rsi 3 on jupiter notebook to improve the workflow reproducibility here i show you a basic data visualization notebook as a demo of automation you can reproduce this demo with this notebook and side escape first launch site escape and then open the notebook in jupyter this notebook imports a protein protein interaction network and gives visual effects according to the quantitative value in addition the notebook experiments with different network layouts and select protein nodes with quantitative values above a certain value and finally expand the selection with the select first neighbors function you might think that this automation is too easy to use however the site escape app and complex workflows are actually a collection of these basic operations getting in the habit of automating usual workflows with ease jupiter notebook not only improves the reproducibility of your research but also gives you ideas for packaging new feature modules i showed an example in good points of automation with jupiter notebook and side escape in fact the combination of jupiter and side escape is not always a good thing there are some side escape features that are not yet provided by the rest api and the environmental setup is a bit difficult for beginners there are several ways to set up a combination of jupiter and side escape one is to install everything in a local environment this is a bit difficult for users as i mentioned before the next one is to install everything in a remote environment this is also a bit problematic this is because site escape does not have a headless mode and requires the setup of a virtual desktop service in a remote environment so a moderate type has been developed by barry demchik the figure on the right shows the combination of remote jupiter and local side escape this is called jupiter bridge it is difficult to open the rest server of a desktop application to the public and allow access to it from the outside so the jupiter bridge replaced the instructions of the pi 4 side escape rest request from remote environment to a request from a local web browser by using the jupiter bridge we can control the local side escape from google collab or binder as follows this is the remote google collab on the right is the local side escape jupiter bridge runs in this cell after running jupiter bridge we can communicate with the local side escape from the remote google collab in this way so far i've been working with side escape outside of jupiter but ideally we should work with side escape not only outside of jupiter notebook but inside of it because there are some side escape features that are not yet provided by the rest api to achieve this i would like to refer to a project called slicer jupiter slicer jupiter is a collection of jupiter kernel and widgets for 3d slicer 3d slicer is a software for visualization and medical image computing 3d slicer is a cross-platform desktop application written in c plus with slicer jupiter we can not only automate the operation of slicer and jupiter but also we can operate slicer interactively in jupiter i think that side escape can use the same approach as 3d slicer to strengthen its collaboration with jupiter here's a summary of what i want to convey jupiter can be used to complement desktop application for biological network visualization it's important for the research reproducibility and jupiter can be used to replace what was previously done in java with python or r this is important for rapid research and development there are multiple local and remote software combinations for working with jupiter notebook and desktop application there is precedent to follow slicer 3d project is developing kernel and widgets for their desktop application to improve the jupiter notebook environment i would like to thank these people this automation project is supported by google season of docs last year in fact at first sight escape automation was more advanced than r than python alex and christina helped with the season of duck's work guitar berry has developed the pi 4 side escape jupiter bridge and the early site escape rest api pi 4 side escape is a port of rsi3's api to python so that it can be used with exactly same usability so i ported what i did first in r to python jorge developed by two side escape it's the predecessor of pi4 side escape k also developed the cytoskay pressed api and pi 2 side escape i would like to thank jupiter and those who provide the notebook service collab and binder i would also like to thank pi data i organized bi data in osaka japan and i got slicers information from the pi data network please refer these papers for more information on site escape automation i would like to thank you all for your attention today