Deep Learning in Genomics Course Material - ALL

This page contains ALL material for the GENE46100 Deep learning in genomics course, including test notes.
Find the 2025 syllabus here.
Edit github source here

Preprocessing and clustering

The data used in this basic preprocessing and clustering tutorial was collected from bone marrow mononuclear cells of healthy human donors and was part of openproblem’s…

Unit 04 Single Cell Genomics

© HakyImLab and Listed Authors - CC BY 4.0 License

Seurat - Guided Clustering Tutorial

gene46100

tutorial

For this tutorial, we will be analyzing the a dataset of Peripheral Blood Mononuclear Cells (PBMC) freely available from 10X Genomics. There are 2,700 single cells that were…

Borzoi prediction from personal genome - qmd version

gene46100

notebook

Using Borzoi for predicting on a personal genome, derived from VCF

Predicting alternative polyadenylation with Borzoi

notebook

gene46100

Borzoi is a deep learning model that, similarily to Enformer:

homework 6

gene46100

homework

If you run the analysis locally, make sure you clone the gene46100 conda environment and install the python packages as needed. See this jupyter notebook for an example.…

Enformer Architecture

gene46100

Enformer is a deep learning model designed for predicting gene expression and other functional genomics signals directly from DNA sequence. It was introduced by DeepMind and…

unit 03

gene46100

Slides for Unit 3 - Enformer, Borzoi, and PrediXcan link

Enformer training

gene46100

notebook

train enformer on human and mouse

Enformer usage neanderthal - jupyter notebook version

gene46100

notebook

predict neanderthal epigenome with enformer

homework 5

gene46100

homework

© HakyImLab and Listed Authors - CC BY 4.0 License

Embeddings Analysis

** embeddings provided by Henry from his DNA GPT model**

homework 4

gene46100

homework

Watch Karpathy’s nanoGPT tutorial on youtube link and answer the questions in the form

Building a GPT - companion notebook annotated

Companion notebook from Karpathys video on building a minimal GPT, annotated by cursors LLM with summary from gemini.

Andrey Karpathy

how to install local llm

how-to

llm

#how-to install ollama open webui local llm on a macbook adapted using conda instead of pyenv from https://medium.com/@hautel.alex2000/build-your-local-ai-from-zero-to-a-cust…

homework 3

gene46100

homework

© HakyImLab and Listed Authors - CC BY 4.0 License

Calibrating hyperparameters with weights and biases

gene46100

project

notebook

Weights and biases is a platform used for AI developers to track, visualize and manage their ML models and experiments. The coolest part is that W&B allows you to log…

TF binding prediction challenge

gene46100

project

Competition details for TF binding prediction challenge

homework 2

gene46100

homework

© HakyImLab and Listed Authors - CC BY 4.0 License

Updated - DNA score prediction with Pytorch

gene46100

notebook

created by Erin Wilson. Downloaded from here.

TF Binding prediction project

gene46100

project

notebook

The goal of this project is to create a neural network model that predicts TF binding strength in a DNA sequence.

Fit linear mode to linear data

We generate a data Y linearly dependent on features X: Y = X\beta + \epsilon

Measuring information in DNA sequence

In the analysis of DNA sequences, particularly when studying motifs and patterns, two key concepts from information theory are entropy and information content. These…

homework 1

gene46100

homework

© HakyImLab and Listed Authors - CC BY 4.0 License

Quick introduction to deep learning

gene46100

notebook

Understand and implement gradient descent to fit a linear model

Boxiang Liu, modified by Haky Im

preparing environment for unit 00

gene46100

how-to

setting up conda environement and installing packages

Building a GPT From Scratch summary

gene46100

slides

gpt

Karpathy’s gpt_dev.ipynb summarized by gemini, reviewed by me.

Gradient descent illustration

© HakyImLab and Listed Authors - CC BY 4.0 License

Conda Environments for Quarto Documents

how-to

How to specify a specific Conda environment when rendering quarto documents.

Borzoi prediction from personal genome - jupyter notebook

gene46100

notebook

Using Borzoi for predicting on a personal genome, derived from VCF

Slides unit 00

An MLP is a fully connected feedforward neural network — the classic deep learning architecture.

TODO

© HakyImLab and Listed Authors - CC BY 4.0 License

© HakyImLab and Listed Authors - CC BY 4.0 License