mapTALL

Transcriptomic subtyping of T-cell acute lymphoblastic leukaemia (T-ALL) using scPred and projection onto the St Jude reference dataset.

---

Overview

mapTALL is an R-based workflow for assigning transcriptomic subtypes to bulk RNA-seq T-ALL samples by projecting them into the feature space defined by the St Jude reference cohort published by Pölönen et al. (2024) and applying supervised classification using scPred.

Although scPred was originally developed for single-cell RNA-seq, mapTALL applies the same framework to bulk RNA-seq by treating each sample as a “cell” in a Seurat object.

The workflow is designed to be:

• transparent
• modular
• conservative in interpretation
• suitable for public reuse

mapTALL is provided as a script-based workflow rather than an R package.

---

Reference data

St Jude reference cohort

The reference dataset is derived from the publicly released transcriptomic data accompanying Pölönen et al. (2024). The full reference object is constructed internally using the authors’ published subtype annotations and dimensional reduction strategy.

For public distribution, a lean reference object is provided:

• contains RNA counts and normalised data for those genes
• retains the PCA reduction defining the scPred feature space
• includes essential annotation fields:
  • Reviewed.subtype
  • Reviewed.genetic.subtype
  • meta_parent

The lean reference object is created after adding a parent classification to the St Jude annotations, followed by removal of bulky components (e.g. scaled data, neighbour graphs) for public distribution.

Reference object

Due to file size constraints, the St Jude reference Seurat object used by mapTALL is not stored directly in this GitHub repository.

The reference object is hosted on the Open Science Framework (OSF):

https://osf.io/adht8/

Users must download the reference file:

• SJ_with_parent.lean.obj.rds

and place it in a local Data/ directory before running the mapTALL script.

---

What the script does

The mapTALL run script performs the following steps:

1. Load inputs

• A pre-built St Jude reference Seurat object
• A user-provided test cohort (bulk RNA-seq counts)

2. Pre-processing

• Intersects test genes with the reference gene set
• Applies ComBat-seq batch adjustment to the shared gene set
• Constructs a Seurat object for the test cohort

3. scPred classification

Three scPred classifiers are applied sequentially:

1. Reviewed.subtype

   • Primary transcriptomic subtype classification
   • Applied to all samples

2. Reviewed.genetic.subtype

   • Applied only to samples whose predicted subtype belongs to a genetic-eligible family

3. meta_parent

   • Higher-level grouping of transcriptomic subtypes

---

Genetic subtype logic

Genetic subtype prediction is conditional on subtype applicability.

Genetic-eligible subtype families

By default, genetic subtype classification is only attempted for samples predicted as:

• TAL1_AB-like
• TAL1_DP-like
• TLX3
• ETP-like
• NKX2-1

This reflects the biological scope of the genetic subtype annotations in the Pölönen dataset.

Interpretation of outputs

Value	Meaning
NA	Genetic subtype is not applicable for this subtype family
unassigned	Genetic subtype is applicable, but confidence threshold was not met
Label	High-confidence genetic subtype assignment

This distinction is intentional and preserved in the output tables.

---

Outputs

The script produces the following key outputs:

1. Summary table

Results_mapTALL_summary.csv

Per sample:

• transcriptomic subtype prediction, no-rejection class, and max probability
• genetic subtype prediction (or NA where not applicable)
• meta-parent classification
• risk category (if enabled)

2. Annotated Seurat object

test.scpred.obj.rds

Contains all predictions and probabilities in @meta.data.

3. Metadata table

test.scpred.meta.csv

Flat table of Seurat metadata for downstream analysis.

4. Session information

A dated sessionInfo_YYYY-MM-DD.txt file for reproducibility.

---

Requirements

• R ≥ 4.2
• Required packages:
  • Seurat (v5)
  • scPred
  • tidyverse
  • data.table
• Optional:
  • GSVA
  • PROGENy

---

Citation

If you use mapTALL, please cite:

Pölönen et al. (2024)
The genomic basis of childhood T-lineage acute lymphoblastic leukaemia

Alquicira-Hernandez et al. (2019)
scPred: accurate supervised method for cell-type classification from single-cell RNA-seq data

mapTALL workflow

If mapTALL or the parent-level classification (meta_parent) introduced in this repository is used in analysis or publication, please cite this repository:

Corley SM. mapTALL: Transcriptomic classification of T-ALL using scPred and the St Jude reference cohort. GitHub repository.
https://github.com/CCICB/mapTALL

A manuscript describing mapTALL is in preparation.

Contact

Questions, issues, or suggestions should be raised via GitHub Issues.