Skip to content

Home

Jump to bottom
jsxlei edited this page May 27, 2019 · 13 revisions

Single-Cell ATAC-seq analysis via Latent feature Extraction

Installation

SCALE neural network is implemented in Pytorch framework.
Running SCALE on CUDA is recommended if available.

install via pip:

pip install scale-atac

install from GitHub

git clone git://github.com/jsxlei/SCALE.git
cd SCALE
python setup.py install

Installation only requieres a few minutes.

Quick Start

Input scATAC-seq data should be one of:

  • dense matrix: data.txt / data.txt.gz
  • sparse matrix: data.mtx / data.mtx.gz, with peaks.txt and cell_id.txt

Run SCALE with provided cluster number k if k is known:

SCALE.py -d [input_dir] -k [k]

Or Run SCALE with estimated k:

SCALE.py -d [input_dir]

Useful options

  • save results in a specific folder: [-o] or [--outdir]
  • filter rare peaks if the peaks quality if not good or too many: [-x]
  • modify the initial learning rate, default is 0.002: [--lr]
  • change the batch size, default is 32: [--batch_size]
  • change iterations by watching the convergence of loss, default is 30000: [-i] or [--max_iter]
  • change random seed for parameter initialization, default is 18: [--seed]
  • binarize the imputation values: [--binary]

Note

If come across the nan loss,

  • try another random seed
  • filter peaks with harsher threshold like -x 4 or -x 6
  • change the initial learning rate to 0.0002

Results

Results will be saved in the output folder including:

  • model.pt
  • feature.txt
  • cluster_assignments.txt
  • imputed_data.txt
  • tsne.txt
  • tsne.pdf

Help

Look for more usage of SCALE

SCALE.py --help

Use functions in SCALE packages.

import scale
from scale import *
from scale.plot import *
from scale.utils import *

Running time

Tutorial

Run SCALE on dense matrix, e.g. Forebrain dataset (k=8, 2088 cells):

SCALE.py -d Forebrain -k 8

Run SCALE on sparse matrix, e.g. Mouse Atlas dataset (k=30, ~80,000 cells):

SCALE.py -d mouse_atlas -k 30 -x 4

Clone this wiki locally