Bioinformatics & Deep Learning — Course Notes
An open-access companion for the redesigned KB8029 course
Welcome
This is an open, interactive companion to the redesigned KB8029 Bioinformatics course. Rather than teaching classical bioinformatics and machine learning as two separate blocks, this book follows the course’s integrated day-by-day structure — see the Course Schedule for the full plan, or course-redesign.md in the repository root for the rationale behind it.
This is not a copy of any textbook. The course previously assigned copyrighted textbooks (Magnus Ekman’s Learning Deep Learning for the machine-learning half, Jonathan Pevsner’s Bioinformatics and Functional Genomics for the classical half). Nothing from either is reproduced here. Every page is written from scratch, drawing on fully open sources — none restricted to non-commercial use — that cover the same material:
- Dive into Deep Learning (Zhang, Lipton, Li & Smola) — CC BY-SA 4.0. Primary source for every machine-learning session.
- Wikipedia — CC BY-SA 4.0. Conceptual and algorithmic depth for the bioinformatics sessions.
- EMBL-EBI Training — CC BY 4.0. The practical, tool-use layer for databases and alignment.
- NCBI Bookshelf — public domain (US government work). BLAST and database practicalities.
- RCSB PDB-101 — CC BY 4.0 (educational tier only). Protein structure.
- A Byte of Python — CC BY-SA 4.0. Python fundamentals and object-oriented programming.
Every page names which source(s) it draws on, and because of the CC BY-SA license, pages adapted from Dive into Deep Learning or Wikipedia are themselves shared here under CC BY-SA 4.0.
Students are never pointed at raw source material directly — Wikipedia in particular is written to be browsed outward via its own links, which makes “how far should I read?” an unanswerable question. Every session gets one bounded, authored page instead, with sources listed for further reading, never as an inline dependency.
How to use this alongside the course
Most session pages have three parts:
- The idea — a short, from-scratch explanation of the session’s core concept, written to match what’s covered in the lecture.
- The notebook — for sessions with a full lab, the actual lab notebook, embedded with its saved output so you can read it here or run it yourself.
- What’s next — an explicit pointer to where this session’s ideas resurface later in the course. The redesign is built around several continuous threads (most centrally, secondary-structure prediction) rather than disconnected topics, and each page names the connection rather than leaving it implicit.
Status
Work in progress. Day 1 (math prerequisites) and Day 7 (linear models, gradient descent, and the XOR problem) are drafted as complete examples — Day 7 in particular showcases the course’s central design idea: machine learning is motivated from inside the bioinformatics material (a profile score turns out to already be a linear model) rather than introduced as an unrelated new topic. The remaining days will follow the same pattern.