Midterm Report

Summary of what happened till the Midterm Evaluation

---

---

The coding period leading up to the Midterm Evaluation has been an amazing experience for me. I’ve organized this report into two distinct parts, with each part dedicated to a specific phase of the project up to the Midterm Evaluation.

Phase 1 ^{30th May - 30th Jun}

• Delved into relevant research papers, understood the intricacies of assembly runs, samples, studies and analyses, which form the foundational concepts of MGnify.

• Documented all visualizations present on the MGnify website which was extensively used in Phase 2.

• Significant chunk of Phase 1 was spent contributing and collaborating with Alejandra Escobar (MGnify Bioinformatician) on the enhancement of her Jupyter notebook - Drawing presence/absence KOs for one metagenomic sample, specifically focusing on Pathways Visualization.

• The objective was to streamline the notebook, making it more concise, interactive, self-explanatory, and immediately usable.
  
  

• Technologies Involved : R + Docker + JupyterLab + Quarto + Packages like KEGGREST, Pathview
  
  

• Pull Request :

  • Worked on an existing PR by Alejandra, Link : https://github.com/EBI-Metagenomics/notebooks/pull/26

  • Status : Under Review 🟡

• Code in Action :

  • Static Version of the Pathways Visualization Notebook I contributed to (hosted on this blog)

  • Documentation of all Visualizations present in MGnify

: : : : : : : : : : : : : : : : : :

Phase 2A ^{1st Jul - 16th Jul}

• This phase focused on finding visualization solutions that could work seamlessly in both JavaScript and Python environments.

• Involved testing of packages like Plotly and Highcharts were tested, with a preference for Highcharts due to its usage in the MGnify site.

• But then Highchart posed an issue as it couldn’t reference data as a link, impacting interoperability between React and Jupyter, especially with large datasets.

• Then came the idea of a common visualization grammar, enabling easy porting of visualizations to Jupyter notebooks for users to perform deeper visualization and push in custom data as well.

• Alternatives like pandas-js, Danfo.js, and Vega-Lite were considered.

• In the end, the Vega ecosystem was selected due to its compatibility and ability to handle large datasets, ensuring seamless interoperability.