mapTALL

Transcriptomic Subtyping of T-ALL by Projection onto the St. Jude Reference Using scPred

mapTALL is an R-based workflow for assigning transcriptomic subtypes to T-cell acute lymphoblastic leukaemia (T-ALL) samples.
It projects local cohort RNA-seq data onto the large St. Jude (SJ) reference dataset published by Pölönen et al. (2024) and performs subtype classification using scPred.

Although scPred was originally designed for single-cell data, mapTALL applies the same supervised-learning approach to bulk RNA-seq, treating each bulk sample as a “cell” within a Seurat object.

mapTALL is a transparent, modular analysis workflow, not a standalone R package.

---

🔍 Key Features

• Uses the curated SJ T-ALL reference (N = 1335) with Reviewed subtype and genetic subtype labels
• Supervised classification via scPred using:
  • 300 variable genes defined by Pölönen
  • 50 principal components
  • Probability-based class assignment
• Multi-level prediction:
  • Reviewed.subtype (e.g., TAL1-DP-like, LMO2/LYL1, TLX1, TLX3, ETP)
  • Reviewed.genetic.subtype (e.g., STIL-TAL1, HOXA-MLL, TLX3-RAG)
  • meta_parent grouping (TAL1, TLX, NKX2, HOXA_MLL, ETP, etc.)
• Built-in confidence metrics:
  • scpred_prediction
  • scpred_no_rejection
  • scpred_max (max probability)
• QC module categorises samples as:
  • High confidence
  • Medium confidence
  • Discordant / Ambiguous
• Optional downstream analyses:
  • TAL1 DP–AB axis scoring
  • GSVA and PROGENy pathway activity
  • Drug-response signatures (e.g., dasatinib sensitivity)

---

📦 Requirements

• R ≥ 4.2
• Required packages:
  • Seurat
  • scPred
  • tidyverse
  • data.table
  • ggplot2, patchwork
• Optional:
  • GSVA
  • progeny

Dataset availability:
The St. Jude reference dataset is not included and must be downloaded from the Pölönen et al. 2024 repository under the appropriate data access conditions.