Spatially resolved, highly multiplexed RNA profiling in single cells

Kok Hao Chen, Alistair N. Boettiger, Jeffrey R. Moffitt, Siyuan Wang, Xiaowei Zhuang

This publication can be found at https://science.sciencemag.org/content/348/6233/aaa6090 and the data referenced below can be downloaded from

Checklist: - [x] point locations - [ ] cell locations - [x] cell x gene expression matrix (derivable)

Load the data

import requests
from io import BytesIO

import pandas as pd

import starspace
from starspace.constants import *

response = requests.get(
    "https://d24h2xsgaj29mf.cloudfront.net/raw/merfish_chen_2015_science_imr90/"
    "140genesData.xlsx"
)
data = pd.read_excel(BytesIO(response.content))

name = "merfish chen 2015 science imr90"

This data file is a cell x gene expression matrix that contains additional metadata as columns of the matrix. Extract those extra columns and clean up the data file.

# map column names to schema

column_map = {
    "RNACentroidX": SPOTS_REQUIRED_VARIABLES.X_SPOT,
    "RNACentroidY": SPOTS_REQUIRED_VARIABLES.Y_SPOT,
    "cellID": "per_slice_cell_id",  # this is not unique experiment-wide
    "CellPositionX": SPOTS_OPTIONAL_VARIABLES.X_REGION,
    "CellPositionY": SPOTS_OPTIONAL_VARIABLES.Y_REGION,
    "geneName": SPOTS_REQUIRED_VARIABLES.GENE_NAME,
    "experiment": "experiment",
    "library": "library",
    "intCodeword": "int_codeword",
    "isCorrectedMatch": "is_corrected_match",
    "isExactMatch": "is_exact_match"
}
columns = [column_map[c] for c in data.columns]
data.columns = columns

# demonstrate that cellID is not unique:
group_columns = (
    "per_slice_cell_id",
    SPOTS_OPTIONAL_VARIABLES.Y_REGION,
    SPOTS_OPTIONAL_VARIABLES.X_REGION,
)

# group by the columns, use size to run a no-op aggregation routine, then drop the size column
# (labeled zero)
not_unique = data.groupby(group_columns).size().reset_index().drop(0, axis=1)

assert_cols = ["per_slice_cell_id"]
assert not_unique[assert_cols].drop_duplicates().shape != not_unique[assert_cols].shape

# fix region ids so that they uniquely identify cells across the experiment.
group_columns = (
    "experiment", "library", "per_slice_cell_id",
    SPOTS_OPTIONAL_VARIABLES.Y_REGION, SPOTS_OPTIONAL_VARIABLES.X_REGION
)
region_ids_map = data.groupby(group_columns).size().reset_index().drop(0, axis=1)

assert_cols = ["per_slice_cell_id", "library", "experiment"]
assert region_ids_map[assert_cols].drop_duplicates().shape == region_ids_map[assert_cols].shape

# map each region to a unique identifier and add it to the data frame
region_ids_map = region_ids_map.drop(
    [SPOTS_OPTIONAL_VARIABLES.Y_REGION, SPOTS_OPTIONAL_VARIABLES.X_REGION], axis=1
)
region_ids_map = region_ids_map.reset_index().set_index(assert_cols)

region_ids = region_ids_map.loc[pd.MultiIndex.from_frame(data[assert_cols])]
data[SPOTS_OPTIONAL_VARIABLES.REGION_ID] = region_ids.values

Write down some important metadata from the publication.

attrs = {
    REQUIRED_ATTRIBUTES.ASSAY: ASSAYS.MERFISH,
    REQUIRED_ATTRIBUTES.SAMPLE_TYPE: "IMR90 lung fibroblast cell line",
    REQUIRED_ATTRIBUTES.AUTHORS: (
        "Kok Hao Chen", "Alistair N. Boettiger", "Jeffrey R. Moffitt", "Siyuan Wang",
        "Xiaowei Zhuang"
    ),
    REQUIRED_ATTRIBUTES.YEAR: 2015,
    REQUIRED_ATTRIBUTES.ORGANISM: "human",
    OPTIONAL_ATTRIBUTES.NOTES: (
        "cellID field from author data renamed per_slice_cell_id to reflect stored data"
    )
}

convert the dataframe into an xarray dataset

spots = starspace.Spots.from_spot_data(data, attrs)

Write the data to zarr on s3

s3_url = "s3://starfish.data.output-warehouse/merfish-chen-2015-science-imr90/"
url = "merfish-chen-2015-science-imr90/"
spots.save_zarr(url)

Convert the xarray dataset to a matrix.

matrix = spots.to_spatial_matrix()
matrix.save_zarr(url)