Skip to content

hakyimlab/deep-learning-genomics

BranchesTags

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
.gitignore
.gitignore
 
 
LICENSE
LICENSE
 
 
README.md
README.md
 
 

Repository files navigation

deep-learning-genomics

Summer immersion syllabus in deep learning in genomics

Organizers/Instructors

  • Hae Kyung Im
  • Ravi Madduri

Students

  • Srusti Donapati - (Uchicago - rising second year College - Katen Scholar 2023)
  • Rachel Rodriguez - (Uchicago - rising second year College - Metcalf fellowship 2023)
  • Sabrina Mi - (UCSD - rising second year Math PhD - R01AA029688)
  • Tiffany Huang - (UChicago - rising third year College - BSD Quantitative Biology Fellowship 2023)
  • Laura Vairus (Harvey Mudd College, rising 2nd year)
  • Dante Vairus (UChicago Lab School - rising 12th year)

Learning Objectives

Syllabus

Draft Syllabus

Welcome to our intensive 8-week course on deep learning in genomics.

In the first 6 weeks, you will learn essential concepts and tools to conduct computational genomic experiments.

For the last 2 weeks, you will conduct a capstone project, where you will apply your newly acquired skills to a real-world project. This involves:

  • Identifying a scientific question
  • Collecting and preprocessing genomic data
  • Conducting experiments and training deep learning models
  • Answering the identified question
  • Presenting the findings to the team
  • Preparing a report following a research paper format

Get ready to explore the fascinating intersection of deep learning and genomics.

The syllabus will change as we will customize it to the needs of the team.

Concepts

  • Central dogma of molecular biology

  • DNA, RNA, proteins

  • transcription, splicing, translation

  • gene regulation

  • GWAS

  • Transcription factors (TFs)

  • Promoters, Enhancers

  • Chromatin structure

  • Histone modifications

  • DNA methylation

  • genetic variation, SNP, structural variation

  • Selection

  • Coalescence and DNA sequence conservation

  • linking variation to function

  • linux file system navigation

  • environmental variables (PATH, SHELL)

Tool list

  • Vscode
  • Code
  • Conda
  • Pip
  • Numpy
  • GitHub
  • Jupyter notebook
  • PyTorch
  • Linux

Cheatsheets

Week 1: Intro and Logistics

Day 1

Day 2

  • Read “LLM in Molecular Biology.” There is a lot of information packed in this survey. Don’t worry if you don’t understand much for now.

  • Discuss how much you understood, make a list of concepts that you did not understand.

  • Create a blog post summarizing what you learned from the article with Quarto and RStudio using this tutorial (you can use ChatGPT) (download Quarto)

  • Publish blog to Quarto Pub

  • What we did

    Read article, made a blog summary of it using Quarto Blog on RStudio and published it with Quarto Pub following tutorials on quarto website

Day 3 June 7, 2023

  • Learn how to create and work with Git and GitHub repositories

  • publish blog from Day 2 on GitHub using this tutorial

  • create a pro user and use co-pilot in github

  • if time:

    • Learn how to use VSCode

  • What we did

    spent the whole time fixing errors on Github, didn’t get to VSCode, recommend having professor more available for questions

git_2x.png

Day 4 June 8, 2023

  • Install VSCode and watch tutorial

  • Learn basics of Python at https://groklearning.com/

  • Install miniconda here (macOS installers)

    • start a blog post on what conda is

  • Python - numpy https://www.w3schools.com/python/numpy/numpy_intro.asp

  • Complete PyTorch tutorials on tensor, autograd, neural networks, and CIFAR10

    basics of matrix algebra using numpy

  • What we did

    Learned basics of using Visual Studio Code.

    Learned and teached each pther basics of python

    Installed conda and environment with much effort and confusion

    • Checkpoints: make a blog on how to use Conda, install conda

      find what path your python is in (which python)

      make env (conda create —name envname)

      activate env (conda activate envname)

      install python in new env (conda install python)

      (conda install pip)

      install python packages with pip (python -m pip install packagename)

      deactivate env (conda deactivate) → returns to default env

    Learned and taught each other basics of python

    split off and practiced python, made tutorial blogs, and started pytorch

    Tiffanie: finished article, matrix algebra exercises

Day 5

  • Learn about OOP (classes, etc.) and dictionaries and files

  • Complete PyTorch tutorials on tensor, autograd, neural networks, and CIFAR10

  • Update Blog with New Page (Day 3/4 materials) & Publish to Github + Quarto

  • Get GitHub Copilot

  • Notes

    • what is PATH? PATH is en env variable in linux (most env var are all caps)
      • shows you path values; order in which it tries to find program
      • if you want to see the value of path write echo $PATH
    • add Code
    • cd / (goes to Macintosh HD all the way to root)
    • exercise: what’s the path to VSCode?
    • for fututre classes of this: make a learning objectives, a preclass quiz, final quiz everyday

  • what we did

    how to install VSCode

    spent a lot of time and effort trying to understand how to download things right.

    Rachel: Classes lesson + exercises, NumPy exercises + NumPy Cheatsheet, VSCode installation and Files system, Files and Dictionaries exercises (VSCode), started PyTorch tutorial

    Laura: made file and dictionary tutorial, did matrix tutorial

    Tiffanie: files review, published LLM summary on github pages, finished matrix algebra exercises, tensor demo

    Srusti: Learned about classes + exercises, NumPy matrix exercises, VSCode installation + files system

Week 2

Day 6

  • Learn what a GWAS is

    • GWAS instructions

      Instructions

      • Read and summarize in a few sentences this GWAS tutorial paper https://onlinelibrary.wiley.com/doi/10.1002/mpr.1608
      • Download plink from LINK (choose the one corresponding to your operating system. If running on posit.cloud, you should choose the linux version even if you are accessing posit from a different operating system)
      • Create a github user if you don’t already have one
      • Git clone the tutorial from the command line (run git clone <https://github.com/MareesAT/GWA_tutorial.git> on the terminal). You will need to unzip files that look like *.zip
      • run the QC and Association components of the tutorial
      ☝🏽 Some modifications to the code may be needed. Use them if you encounter errors.
      • Notes

      Relatedness.R needs to change lines 7 and 15 to

      legend(1,1, xjust=1, yjust=1, legend=levels(factor(relatedness$RT)), pch=16, col=c(4,3))
legend(0.02,1, xjust=1, yjust=1, legend=levels(factor(relatedness$RT)), pch=16, col=c(4,3))

      line 31 of 2_Main_script_MDS.txt replace

      plink --bfile ALL.2of4intersection.20100804.genotypes --set-missing-var-ids @:#[b37]\$1,\$2 --make-bed --out ALL.2of4intersection.20100804.genotypes_no_missing_IDs

      with

      plink --bfile ALL.2of4intersection.20100804.genotypes --set-missing-var-ids '@:#[b37]$1,$2' --make-bed --out ALL.2of4intersection.20100804.genotypes_no_missing_IDs

      line 144 in 2_Main_script_MDS.txt replace

      (base) haekyungim@Im-Lab-016 1_QC_GWAS % cat race_1kG14.txt racefile_own.txt | sed -e '1i\FID IID race' > racefile.txt
sed: 1: "1i\FID IID race
": extra characters after \ at the end of i command
(base) haekyungim@Im-Lab-016 1_QC_GWAS % cat race_1kG14.txt racefile_own.txt | sed -e '1i\
FID IID race' > racefile.txt

      install qqman in R and comment out first line on Manhattan_plot.R and QQ_plot.R

      ##install.packages("qqman",repos="http://cran.cnr.berkeley.edu/",lib="~" ) # location of installation can be changed but has to correspond with the library location 
##library("qqman",lib.loc="~")  
library("qqman")

      Questions

      Using the output from the tutorial or using the commands you learned from it, answer the following questions. Show the command you used to create the result.

      1. How many individuals are in the genotype file you downloaded? (5 pts)
      2. Explain the contents of .fam, .bim, ``.bed files (5 pts)
      3. Write the captions for the pdf's generated by the commands in 1_Main_script_QC_GWAS.txt and 3_Main_script_association_GWAS (5 pts per figure caption)
      4. Discuss what you did and the results you obtained. (20 pts)

    • Run a GWAS, use these instructions

    https://bios25328.hakyimlab.org/post/2021/04/09/lab-2-gwas-in-practice/

  • Start Karpathy's "Zero to Hero" course

    Syllabus zero to hero

  • installing plink (use zsh for your terminal)

    • download plink online https://www.cog-genomics.org/plink/
    • move plink file (not folder, you can leave the rest of the stuff) to bin folder (make a folder called bin in your home directory)
    • stand in that bin in terminal
    • make plink executable: chmod +x plink
    • if it doesn’t let you open it just go to the plink in Finder
    • `nano ~/.zshrc'
    • more
    • less
    • cat
    • only put programs in bin

  • Questions

    What does the tutorial (1_QC_GWAS) mean when it says “cases” in the following instruction: “This second HWE step only focusses on cases because in the controls all SNPs with a HWE p-value < hwe 1e-6 were already removed”?

  • what we did

    Rachel: installed plink, worked on main script QC GWAS tutorial (genetic QC)

Day 7

Day 8

  • Take quantatative genomic training class
  • Learn how to login to the HPC cluster
  • Learn how to run enformer usage and training notebooks
  • Learn how to visualize data with Python's Matplotlib library

Day 9

  • Learn how to visualize data with R's ggplot library
  • Learn basics of Quarto blogging
  • Create a new blog post

Day 10

  • Complete "LLM in Molecular Biology" article
  • Take Deep Learning for Genomics quiz
  • Learn how to summarize research papers
  • Learn how to create multiple choice questions from research papers

Week 3

Day 11

  • Run Temi's pipeline (PrediXcan 1.7)
  • Train predictors of new tracks (TF binding, single cell expression)
  • Learn how to interpret enformer results, models, and attention

Day 12

Day 13

Day 14

  • Create post-lesson Google Forms

Day 15

  • Learn how to create and work with Jupyter Notebooks
  • Learn how to use Pandas for data manipulation

Week 4

Day 16

  • Learn how to use Seaborn for data visualization
  • Complete Seaborn tutorial on Kaggle
  • Learn the basics of statistical analysis

Day 17

  • Learn how to use Weights & Biases for experiment tracking

Day 18

Day 19

Day 20

Week 5

Day 21

Day 22

Day 23

Day 24

Day 25

Week 6

Day 26

Day 27

Day 28

Day 29

Day 30

Select capstone project

  • predict methylation from DNA sequence

  • train personalized Enformer

  • setup llamaindex to allow summarization and question answering with custom text or papers

  • reproduce/implement scGPT

  • train additional epigenetic features

  • run https://github.com/kundajelab/bpnet and compare to TFPred

  • analyze TF binding matrix predicted by Enformer, interpret, attention links?

  • visualize Enformer output to facilitate interpretation

  • (Dante) predict epigenome in Neanderthals

  • (Sabrina) train predictors of rat transcriptome

Week 7 (Capstone Project)

Day 31

Day 32

Day 33

Day 34

Day 35

Week 8 (Capstone Project cont.)

Day 36

Day 37

Day 38

Day 39

Day 40

https://hakyimlab.notion.site/Homework-3-run-a-GWAS-685b2d3b16e1485d913500739a99eb58?pvs=4

About

Summer immersion syllabus in deep learning in genomics

Resources

Readme

License

MIT license
Activity
Custom properties

Stars

1 star

Watchers

3 watching

Forks

1 fork
Report repository

Releases

No releases published

Packages

No packages published