csvranger

A Rust library for parsing the output CSV files of 10x Genomics *ranger pipelines.

Many of 10x Genomics data-processing pipelines produce CSV files that summarize the data. These CSVs, while human-readable, cannot be parsed by a machine without extra effort. This small crate provides the necessary functionality to parse the values in these files. See documentation at csvranger.

Example usage

Using TenxCsvValue::from_csv_value

use csvranger::TenxCsvValue;

fn main() {
    // Sample data from test-data/cellranger_multi.10.0
    let raw_csv = b"Sample ID,Sample barcodes,Sample description,GEX: Cells,GEX: Confidently mapped reads in cells,GEX: Median UMI counts per cell,GEX: Median genes per cell,GEX: Total genes detected
    SOD1_G93A_mouse_spinal_cord_P112_specimen_1,,SOD1-G93A mouse spinal cord from P112 mouse,16410,0.6847850295990157,2244,1305,27219";

    let mut reader = csv::Reader::from_reader(&raw_csv[..]);
    for line in reader.records() {
        // Don't actually call .unwrap in production!
        let line = line.unwrap();
        for value in line.iter() {
            // This is where the magic happens
            TenxCsvValue::from_csv_value(value);
        }
    }
}

Deserializing directly

use csvranger::TenxCsvValue;
use std::collections::HashMap;

fn main() {
    // Sample data from test-data/cellranger_multi.10.0
    let raw_csv = b"Sample ID,Sample barcodes,Sample description,GEX: Cells,GEX: Confidently mapped reads in cells,GEX: Median UMI counts per cell,GEX: Median genes per cell,GEX: Total genes detected
    SOD1_G93A_mouse_spinal_cord_P112_specimen_1,,SOD1-G93A mouse spinal cord from P112 mouse,16410,0.6847850295990157,2244,1305,27219";

    let mut reader = csv::Reader::from_reader(&raw_csv[..]);
    let mut parsed_data: Vec<HashMap<String, TenxCsvValue>> = Vec::new();
    for deserialized_record in reader.deserialize() {
        parsed_data.push(deserialized_record.unwrap())
    }
}

Features

• legacy: If you are parsing CSV-files from legacy 10x Genomics pipelines, you'll need to activate this feature.
• serde: Support for serializing a TenxCsvValue using serde. Note that if the legacy feature is enabled, this will deserialize using TenxCsvValue::from_legacy_csv_value, which may use a small regular expression to extract numerical values from the data. If this behavior is surprising and undesired, please file an issue because it's not a decision I'm 100% sold-on.
• schemars: Support for schemars

You can add features to your project directly to your Cargo.toml:

csvranger = { version = "0.1.0", features = ["legacy"] }

or using cargo:

cargo add csvranger --features legacy

Tested pipelines

The outputs of the following pipeline-version combinations are tested, though others likely work as well:

• cellranger 6
• cellranger 8
• cellranger 9
• cellranger 10
• celranger-atac 2
• spaceranger 4

More pipeline-version combinations will be added.

Legacy pipelines

If you are parsing CSVs produced by cellranger < 10, you'll want to activate the legacy feature. The outputs of cellranger-atac < 2 and spaceranger < 4 are untested, but they likely do not require the legacy feature.