# The functional landscape of alternative splicing in hematopoietic lineage commitment

## TL;DR

This study uses a machine-learning model to identify functional alternative splicing events during blood cell development in vertebrates.

## Contribution

A novel machine-learning model, FAScore, is developed to predict functional exon-skipping events in hematopoietic lineage commitment.

## Key findings

- Four previously-uncharacterized functional AS events were identified that affect erythropoiesis and myelopoiesis.
- Deletion of TBC1D23 exon 15 reduces erythropoiesis in mice and zebrafish by increasing SUMOylation of HDAC1.
- FAScore provides a valuable tool for identifying functional exon skipping during hematopoietic lineage commitment.

## Abstract

Alternative splicing (AS) is a ubiquitous post-transcriptional regulatory mechanism, that has greatly expanded the transcriptomic and proteomic diversity in vertebrates. While gene regulation of hematopoiesis has been extensively researched in vertebrates, the functions of species- and cell lineage-specific splice variants in vertebrates are largely unknown. Here, we curate transcriptomic data on fetal hematopoietic organ development in six vertebrates and hematopoietic cell differentiation in humans and mice. To identify functional exon-skipping events among thousands of cassette exons in protein-coding genes for a specific differentiation lineage and species, we develop a machine-learning model interrogating 19 features including dynamic expression, protein structure, and evolutionary conservation, and integrate them into a single prediction score, named Functional AS Score (FAScore). Using FAScore, we identify four previously-uncharacterized functional AS events in which deletion of the AS exon leads to defects in erythropoiesis and myelopoiesis. Furthermore, we demonstrate that deletion of exon 15 of TBC1D23 reduces erythropoiesis in mice and zebrafish through elevated binding capacity to RANBP2/RANGAP1 leading to increased SUMOylation level of HDAC1. Collectively, our study presents a valuable tool to identify functional exon skipping (ES) events during hematopoietic lineage commitment, and establishes a research paradigm that can be broadly applied to other biological processes.

Alternative splicing greatly expands the transcriptomic and proteomic diversity in vertebrates. Here, the authors develop a machine-learning tool to identify functional splicing events during hematopoietic cell differentiation, revealing new regulators of hematopoiesis.