{ "cells": [ { "cell_type": "markdown", "id": "cc74719d-0dab-4858-8153-5a388750b4d6", "metadata": {}, "source": [ "# Explore Multiome Data\n", "\n", "This is the data exploration notebook for the multimodal data from the Neurips 2021 Multimodal Data Integration Challenge. You can find full documentation for the competition at https://openproblems.bio/neurips_docs/\n", "\n", "The goal for this notebook is to introduce you to the multiome gene expression (GEX) and Assay for Transposase-Accessible Chromatin (ATAC) datasets. " ] }, { "cell_type": "markdown", "id": "8c0762d9-f00c-4a27-a3e8-b778b38174be", "metadata": {}, "source": [ "## Loading the data" ] }, { "cell_type": "code", "execution_count": 17, "id": "0b616a17-980c-4d18-8b83-53d455cd9663", "metadata": { "execution": { "iopub.execute_input": "2021-09-10T16:46:07.490236Z", "iopub.status.busy": "2021-09-10T16:46:07.489636Z", "iopub.status.idle": "2021-09-10T16:46:08.962676Z", "shell.execute_reply": "2021-09-10T16:46:08.961909Z", "shell.execute_reply.started": "2021-09-10T16:46:07.490157Z" }, "tags": [] }, "outputs": [], "source": [ "import numpy as np\n", "import pandas as pd\n", "import logging\n", "from sklearn.decomposition import TruncatedSVD\n", "from sklearn.linear_model import LinearRegression\n", "from sklearn.metrics import mean_squared_error\n", "\n", "import scanpy as sc\n", "import anndata as ad\n", "import matplotlib.pyplot as plt\n", "from datasets import load_dataset, Dataset\n", "from huggingface_hub import hf_hub_download" ] }, { "cell_type": "markdown", "id": "c46128bf-b348-4693-b8ac-7c9b79c06e95", "metadata": {}, "source": [ "The data is stored in two AnnData objects, one for each modality. Full documentation for AnnData is [here](https://anndata.readthedocs.io/en/latest/). \n", "\n", "Let's load the multiome data." ] }, { "cell_type": "code", "execution_count": 18, "id": "8d911781-b881-49fe-9dcc-5c043ec3a16b", "metadata": { "execution": { "iopub.execute_input": "2021-09-10T16:46:08.966278Z", "iopub.status.busy": "2021-09-10T16:46:08.965870Z", "iopub.status.idle": "2021-09-10T16:46:24.275522Z", "shell.execute_reply": "2021-09-10T16:46:24.274638Z", "shell.execute_reply.started": "2021-09-10T16:46:08.966254Z" }, "tags": [] }, "outputs": [ { "data": { "text/plain": [ "AnnData object with n_obs × n_vars = 42492 × 13431\n", " obs: 'pct_counts_mt', 'n_counts', 'n_genes', 'size_factors', 'phase', 'cell_type', 'pseudotime_order_GEX', 'batch', 'pseudotime_order_ATAC', 'is_train'\n", " var: 'gene_ids', 'feature_types', 'genome'\n", " uns: 'dataset_id', 'organism'\n", " obsm: 'X_pca', 'X_umap'\n", " layers: 'counts'" ] }, "execution_count": 18, "metadata": {}, "output_type": "execute_result" } ], "source": [ "REPO_ID = \"paupaiz/Bone_Marrow_BMMCs\"\n", "gex = \"multiome_gex_processed_training.h5ad\"\n", "atac = \"multiome_atac_processed_training.h5ad\"\n", "\n", "adata_gex = ad.read_h5ad(\n", " hf_hub_download(repo_id=REPO_ID, filename=gex, repo_type=\"dataset\"))\n", "adata_gex" ] }, { "cell_type": "code", "execution_count": 19, "id": "987f41fe", "metadata": {}, "outputs": [ { "data": { "text/plain": [ "AnnData object with n_obs × n_vars = 42492 × 116490\n", " obs: 'nCount_peaks', 'atac_fragments', 'reads_in_peaks_frac', 'blacklist_fraction', 'nucleosome_signal', 'cell_type', 'pseudotime_order_ATAC', 'batch', 'pseudotime_order_GEX', 'is_train'\n", " var: 'feature_types'\n", " uns: 'dataset_id', 'gene_activity_var_names', 'organism', 'sample_pm_varnames'\n", " obsm: 'gene_activity', 'lsi_full', 'lsi_red', 'umap'\n", " layers: 'counts'" ] }, "execution_count": 19, "metadata": {}, "output_type": "execute_result" } ], "source": [ "adata_atac = ad.read_h5ad(\n", " hf_hub_download(repo_id=REPO_ID, filename=atac, repo_type=\"dataset\"))\n", "adata_atac" ] }, { "cell_type": "markdown", "id": "e502bf40-ec3d-47c8-88ba-5c814f7e4f86", "metadata": {}, "source": [ "### What is the data?\n", "\n", "Data in AnnData object is stored in `adata.X`. The shape of the data is accessible using `adata.shape` or `adata.[n_obs|n_vars]`" ] }, { "cell_type": "code", "execution_count": 20, "id": "9ffedbaf-3a85-4cdb-a4fe-db83374336c8", "metadata": { "execution": { "iopub.execute_input": "2021-09-10T16:46:24.277944Z", "iopub.status.busy": "2021-09-10T16:46:24.277714Z", "iopub.status.idle": "2021-09-10T16:46:24.283550Z", "shell.execute_reply": "2021-09-10T16:46:24.282742Z", "shell.execute_reply.started": "2021-09-10T16:46:24.277916Z" }, "tags": [] }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "The GEX data has 42492 observations and 13431 features.\n", "The ATAC data has 42492 observations and 116490 features.\n" ] } ], "source": [ "print(f\"The GEX data has {adata_gex.n_obs} observations and {adata_gex.n_vars} features.\")\n", "print(f\"The ATAC data has {adata_atac.n_obs} observations and {adata_atac.n_vars} features.\")" ] }, { "cell_type": "markdown", "id": "feb0b17c-42b5-4b05-9e4a-bd015c9f80c9", "metadata": {}, "source": [ "As we can see, there are many more features in the ATAC matrix than in the GEX matrix. This is expected, because ATAC measures accessibility over the entire genome, while GEX only measures expression for the 25,000 genes in the genome. This data has already been preprocessed, so peaks found in fewer than 15 cells and genes detected in fewer than 20 cells were excluded." ] }, { "cell_type": "markdown", "id": "00a9141d-5b78-4046-bbba-db271a004b73", "metadata": {}, "source": [ "### Preprocessing of gene expression\n", "\n", "In this dataset, gene expression was measured using 3' capture of nuclear RNA as described in the [10X Multiome Product Guide](https://www.10xgenomics.com/products/single-cell-multiome-atac-plus-gene-expression). Note, not all RNA is found in the nucleus. Comparisons of nuclear and cytosolic RNA have been previously reported (e.g. [Bakken 2018](https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0209648); [Abdelmoez 2018](https://genomebiology.biomedcentral.com/articles/10.1186/s13059-018-1446-9)) as have comparisons of single-nucleus and single-cell RNA sequencing ([Lake 2017](https://www.nature.com/articles/s41598-017-04426-w)).\n", "\n", "For gene expression data, cells were filtered based on mitochondrial content, UMI counts per cell, and genes detected per cell. Size factors were then calculated using [scran](https://genomebiology.biomedcentral.com/articles/10.1186/s13059-016-0947-7) and stored in `adata.obs[\"size_factors\"]`. \n", "\n", "Counts were then normalized per cell by divided the UMI counts by the size factors. Original counts are stored in `adata.layers[\"counts\"]`. The size factor normalized counts are stored in `adata.X`.\n", "\n", "Finally, normalized counts are [log1p transformed](https://scanpy.readthedocs.io/en/stable/generated/scanpy.pp.log1p.html). These normalized counts are stores in `adata.layers[\"log_norm\"]`.\n", "\n", "More information about best practices for single-cell analysis can be found [here](https://www.embopress.org/doi/full/10.15252/msb.20188746).\n" ] }, { "cell_type": "markdown", "id": "e3e5316e-4e7b-4568-a0db-2f99ae278b0e", "metadata": {}, "source": [ "### Preprocessing of ATAC\n", "\n", "The chromatin accessibility data acquired by ATAC-seq as part of the 10X Multiome protocol was processed using [Signac](https://satijalab.org/signac/). Quality control, dimensionality reduction and translating peaks to gene activity scores was performed using Signac, following the authors' instructions. After loading the peak-by-cell matrix, counts were binarized to only represent an accessible versus non-accessible state of each region. Cells were then filtered based on 5 quality control metrics comprising the total number of fragments, the enrichment of fragments detected at transcription start sites (TSS), the fraction of fragments in peak regions compared to peak-flanking regions, the fraction of peaks blacklisted by the ENCODE consortium, and the nucleosome signal, which describes the length distribution of fragments which is expected to follow the length of DNA required span across one nucleosome or multiples of it.\n", "\n", "Since ATAC data is sparser than gene expression data, peaks were included if they were accessible in at least 15 cells.\n", "\n", "Finally, the data was binarized by setting all values `>0` to `1`. Raw UMI counts for each peak can be found in `adata.layers[\"counts\"]`." ] }, { "cell_type": "markdown", "id": "c3c42443-d686-4f68-b301-cc3607752065", "metadata": {}, "source": [ "### Feature metadata\n", "\n", "More information about the features are available in the `.var` DataFrame of each object. \n", "\n", "#### GEX feature metadata\n", "\n", "The GEX `adata.var` DataFrames have the following columns:\n", "\n", "* `.var.index` - [Ensembl Gene Names](https://m.ensembl.org/info/genome/genebuild/gene_names.html) for each gene\n", "* `.var[\"gene_ids\"]` - [Ensembl Stable IDs](https://useast.ensembl.org/info/genome/stable_ids/index.html) used to uniquely track genes whose Gene Names may change over time.\n", "* `.var[\"feature_types\"]` - Denotes the each feature as a gene expression feature. Should be `GEX` for all genes\n", "* `.var[\"genome\"]` - The [Genome Assembly](https://www.ncbi.nlm.nih.gov/assembly/GCF_000001405.26/) used for read mapping.\n" ] }, { "cell_type": "code", "execution_count": 21, "id": "3e39bc43-f815-4ca3-8099-fa92363c6d5d", "metadata": { "execution": { "iopub.execute_input": "2021-09-10T16:46:24.285280Z", "iopub.status.busy": "2021-09-10T16:46:24.284905Z", "iopub.status.idle": "2021-09-10T16:46:24.333422Z", "shell.execute_reply": "2021-09-10T16:46:24.332700Z", "shell.execute_reply.started": "2021-09-10T16:46:24.285252Z" } }, "outputs": [ { "data": { "text/html": [ "
\n", " | gene_ids | \n", "feature_types | \n", "genome | \n", "
---|---|---|---|
AL627309.5 | \n", "ENSG00000241860 | \n", "GEX | \n", "GRCh38 | \n", "
LINC01409 | \n", "ENSG00000237491 | \n", "GEX | \n", "GRCh38 | \n", "
LINC01128 | \n", "ENSG00000228794 | \n", "GEX | \n", "GRCh38 | \n", "
NOC2L | \n", "ENSG00000188976 | \n", "GEX | \n", "GRCh38 | \n", "
KLHL17 | \n", "ENSG00000187961 | \n", "GEX | \n", "GRCh38 | \n", "
... | \n", "... | \n", "... | \n", "... | \n", "
MT-ND5 | \n", "ENSG00000198786 | \n", "GEX | \n", "GRCh38 | \n", "
MT-ND6 | \n", "ENSG00000198695 | \n", "GEX | \n", "GRCh38 | \n", "
MT-CYB | \n", "ENSG00000198727 | \n", "GEX | \n", "GRCh38 | \n", "
AL592183.1 | \n", "ENSG00000273748 | \n", "GEX | \n", "GRCh38 | \n", "
AC240274.1 | \n", "ENSG00000271254 | \n", "GEX | \n", "GRCh38 | \n", "
13431 rows × 3 columns
\n", "\n", " | feature_types | \n", "
---|---|
chr1-9776-10668 | \n", "ATAC | \n", "
chr1-180726-181005 | \n", "ATAC | \n", "
chr1-181117-181803 | \n", "ATAC | \n", "
chr1-191133-192055 | \n", "ATAC | \n", "
chr1-267562-268456 | \n", "ATAC | \n", "
... | \n", "... | \n", "
GL000219.1-90062-90937 | \n", "ATAC | \n", "
GL000219.1-99257-100160 | \n", "ATAC | \n", "
KI270726.1-27152-28034 | \n", "ATAC | \n", "
KI270713.1-21434-22336 | \n", "ATAC | \n", "
KI270713.1-29629-30491 | \n", "ATAC | \n", "
116490 rows × 1 columns
\n", "\n", " | pct_counts_mt | \n", "n_counts | \n", "n_genes | \n", "size_factors | \n", "phase | \n", "cell_type | \n", "pseudotime_order_GEX | \n", "batch | \n", "pseudotime_order_ATAC | \n", "is_train | \n", "
---|---|---|---|---|---|---|---|---|---|---|
TAGTTGTCACCCTCAC-1-s1d1 | \n", "1.061008 | \n", "1508.0 | \n", "1022.0 | \n", "0.453484 | \n", "S | \n", "Naive CD20+ B | \n", "NaN | \n", "s1d1 | \n", "NaN | \n", "True | \n", "
CTATGGCCATAACGGG-1-s1d1 | \n", "0.604230 | \n", "1655.0 | \n", "1081.0 | \n", "0.455631 | \n", "G2M | \n", "CD14+ Mono | \n", "NaN | \n", "s1d1 | \n", "NaN | \n", "True | \n", "
CCGCACACAGGTTAAA-1-s1d1 | \n", "0.650069 | \n", "7230.0 | \n", "3304.0 | \n", "2.435348 | \n", "G2M | \n", "CD8+ T | \n", "NaN | \n", "s1d1 | \n", "NaN | \n", "True | \n", "
TCATTTGGTAATGGAA-1-s1d1 | \n", "0.812274 | \n", "1108.0 | \n", "793.0 | \n", "0.347226 | \n", "G2M | \n", "CD8+ T | \n", "NaN | \n", "s1d1 | \n", "NaN | \n", "True | \n", "
ACCACATAGGTGTCCA-1-s1d1 | \n", "1.674770 | \n", "1851.0 | \n", "1219.0 | \n", "0.534205 | \n", "G2M | \n", "CD16+ Mono | \n", "NaN | \n", "s1d1 | \n", "NaN | \n", "True | \n", "
... | \n", "... | \n", "... | \n", "... | \n", "... | \n", "... | \n", "... | \n", "... | \n", "... | \n", "... | \n", "... | \n", "
ATTCACTTCCTGCGAA-14-s3d7 | \n", "0.000000 | \n", "1475.0 | \n", "967.0 | \n", "0.636243 | \n", "G2M | \n", "CD8+ T | \n", "NaN | \n", "s3d7 | \n", "NaN | \n", "True | \n", "
GCTCTGTTCTGCAAGT-14-s3d7 | \n", "0.014362 | \n", "6963.0 | \n", "3237.0 | \n", "3.019859 | \n", "S | \n", "G/M prog | \n", "NaN | \n", "s3d7 | \n", "NaN | \n", "True | \n", "
GCTGAGGAGTGAGCGG-14-s3d7 | \n", "0.068399 | \n", "1462.0 | \n", "876.0 | \n", "0.495807 | \n", "G2M | \n", "Erythroblast | \n", "0.791014 | \n", "s3d7 | \n", "0.828536 | \n", "True | \n", "
TACTGAGGTTCGCTCA-14-s3d7 | \n", "0.000000 | \n", "2808.0 | \n", "1584.0 | \n", "1.125766 | \n", "G2M | \n", "CD14+ Mono | \n", "NaN | \n", "s3d7 | \n", "NaN | \n", "True | \n", "
CAATAAGCAGATAGAC-14-s3d7 | \n", "0.000000 | \n", "1488.0 | \n", "970.0 | \n", "0.636224 | \n", "S | \n", "CD4+ T naive | \n", "NaN | \n", "s3d7 | \n", "NaN | \n", "True | \n", "
42492 rows × 10 columns
\n", "\n", " | nCount_peaks | \n", "atac_fragments | \n", "reads_in_peaks_frac | \n", "blacklist_fraction | \n", "nucleosome_signal | \n", "cell_type | \n", "pseudotime_order_ATAC | \n", "batch | \n", "
---|---|---|---|---|---|---|---|---|
AAACAGCCAACACTTG-2-s1d2 | \n", "6421.0 | \n", "8963 | \n", "0.716390 | \n", "0.000623 | \n", "0.832734 | \n", "NK | \n", "NaN | \n", "s1d2 | \n", "
AAACAGCCAATAGTCT-s2d4 | \n", "3232.0 | \n", "2637 | \n", "1.225635 | \n", "0.000000 | \n", "0.888535 | \n", "Naive CD20+ B | \n", "NaN | \n", "s2d4 | \n", "
AAACAGCCAATTAAGG-1-s1d1 | \n", "4624.0 | \n", "6558 | \n", "0.705093 | \n", "0.001730 | \n", "0.688525 | \n", "CD4+ T naive | \n", "NaN | \n", "s1d1 | \n", "
AAACAGCCACCAGGTT-2-s1d2 | \n", "7177.0 | \n", "9577 | \n", "0.749400 | \n", "0.000557 | \n", "0.772973 | \n", "NK | \n", "NaN | \n", "s1d2 | \n", "
AAACAGCCAGCTACGT-s2d4 | \n", "5269.0 | \n", "4573 | \n", "1.152198 | \n", "0.000380 | \n", "0.846552 | \n", "CD8+ T | \n", "NaN | \n", "s2d4 | \n", "
... | \n", "... | \n", "... | \n", "... | \n", "... | \n", "... | \n", "... | \n", "... | \n", "... | \n", "
TTTGTTGGTGTAATAC-s2d4 | \n", "905.0 | \n", "891 | \n", "1.015713 | \n", "0.000000 | \n", "1.080808 | \n", "Erythroblast | \n", "0.697933 | \n", "s2d4 | \n", "
TTTGTTGGTTAGACCA-2-s1d2 | \n", "6669.0 | \n", "8699 | \n", "0.766640 | \n", "0.001799 | \n", "0.639382 | \n", "CD8+ T | \n", "NaN | \n", "s1d2 | \n", "
TTTGTTGGTTCGGTAA-1-s1d1 | \n", "7074.0 | \n", "9873 | \n", "0.716500 | \n", "0.001414 | \n", "0.868516 | \n", "NK | \n", "NaN | \n", "s1d1 | \n", "
TTTGTTGGTTTGGGTA-4-s2d1 | \n", "8404.0 | \n", "13557 | \n", "0.619901 | \n", "0.000952 | \n", "1.495506 | \n", "CD4+ T naive | \n", "NaN | \n", "s2d1 | \n", "
TTTGTTGGTTTGGTTC-s2d4 | \n", "1437.0 | \n", "1256 | \n", "1.144108 | \n", "0.001392 | \n", "1.185714 | \n", "cDC2 | \n", "NaN | \n", "s2d4 | \n", "
20952 rows × 8 columns
\n", "