Manifest

Manifest purpose

The RAFT manifest defines the relationships among samples, patients, and datasets. The number and types of samples required for each RAFT workflow varies. For example, an RNA quantification workflow may use one or multiple RNA-sequencing samples per patient. Other workflows, like somatic variant calling, will require both a normal sample and a tumor sample for each patient.

RAFT manifests may contain one or more patients. Many computer clusters will allow for multiple patients to be run in parallel to reduce run time.

Sample, patient, and dataset heirarchy

The general hierarchy of organization within RAFT follows

Sample ∈ Patient ∈ Dataset

In other words, samples belong to patients (patients can have multiple samples) and patients belong to datasets (datasets can have multiple patients).

Manifest contents

A RAFT manifest must have at least the columns defined in the table below. Columns can be in any order and other columns containing non-RAFT metadata are also allowed.

RAFT Columns
Column	Description	Allowed values
Dataset	Name for collection of patients	Free text
Patient Name	Name for collection of samples	Free text
Run Name	Name for the specific sample	Free text (see note below)
File Prefix	Base name (or full path) of input files	Free text
Sequencing Method	Sequencing protocol for sample	(RNA-Seq, WES, WXS, WGS)
Normal	Is the sample normal or abnormal (tumor)?	(TRUE, FALSE)

Note

A sample’s Run_Name is instrumental in guiding samples through some RAFT workflows. A sample’s Run_Name should have a two-letter prefix that describes the type of sample followed by an arbitrary unique identifier. The first letter of the prefix is either a (for abnormal) or n (for normal). The second letter is either r (for RNA) or d (for DNA). For example, a sample with an ar- prefix is an abnormal (tumor) RNA sample while a sample with a nd- prefix is a normal DNA sample.

Each line in the manifest after the header corresponds to a sample and provides the necessary data for running a RAFT workflow. The samples described within the manifest may, in some cases, be effectively independent (as in, the workflow does not attempt to pair samples from a patient), but in other cases, users must be careful that samples are properly labeled. For example, somatic variant calling generally requires a normal DNA sample and a tumor DNA sample. For RAFT to properly pair these samples together, they must have the correct sample prefix (nd- for the DNA tumor sample and nd- for the DNA normal sample) and be paired with the patient (Patient_Name field) and dataset (Dataset field). Consider the following example:

Patient_Name Run_Name Dataset File_Prefix Sequencing_Method Normal
Pt01  ad-Pt01-03A     AML     9f7f7   WES     FALSE
Pt01  nd-Pt01-11A     AML     8e74a   WES     TRUE
Pt01  ar-Pt01-03A     AML     cdb288  RNA-Seq FALSE
CTRL  nr-CTRL         AML     CD34-U  RNA-Seq TRUE

Note

Both the tumor DNA sample (ad-Pt01-03A) and the normal DNA (nd-Pt01-11A) sample belong to the same patient (Pt01) and the same dataset (AML).

User-provided manifests can be sanity-checked by using:

raft.py check-manifest -m </PATH/TO/MANIFEST>

Users can then provide the manifest to RAFT’s run-ots command to run the workflow with their manifest.