To evaluate protein-protein interaction (PPI) experiment results, I’ve searched 4 PPI database as a cross-reference. Here’s a brief summary on the 4 databases.

230511: Reading in progress.

The pre-trained attention-based language models have proved themselves super powerful. In life science field, models pre-trained on protein sequences can also extract a considerable amount of hidden information. A short note on BERTology in this blog.

Scenario: we have data from multiple experiments. The experiment names are joint by project name and auto-incremented task id. The number of records in each task varies, and the number of tasks in each project also varis. We want to plot a boxplot to examine the distribution but would like to color boxes from same project as the same. To do so we need to set the palette argument.

To make the version control more efficient, we don’t want git to track the output of jupyter notebooks every time a notebook is committed. If the output is indeed necessary we can always convert an HTML as a report and keep it static.

Deploy to Rstudio Connect from Git

It’s been a long time since I last updates the blog. A lot of things have happened. Time to carry on.

This work is an OPIG (Oxford Protein Informatics Group) project, like SAbDab.

Great demonstration of model interpretation from Salesforce research team. Project in Github. From the project name “provis” we can get of hint of the authors’ initial design.

Evernote is great. I kept tech notes in Evernote from 10 years ago. That include learning notes, literature reviews, and tech tips. The notes marked my transition from a fresh bachelor in biology to a bioinformatician, and machine learning / data science player. Within this blog I’m going to migrate some recent notes and most importantly synchronize future notes.