Pangenomes, cell atlases and CRISPR's first US approval

The graph replaces the line

On 11 May 2023 the Human Pangenome Reference Consortium published 'A draft human pangenome reference' in Nature, volume 617. The draft comprises 47 phased, diploid assemblies from genetically diverse individuals, covering more than 99 percent of expected sequence per genome at better than 99 percent accuracy at both structural and base-pair level.

The computational consequence is substantial. A graph reference is not a drop-in replacement for GRCh38; alignment, variant calling, coordinate systems and every downstream annotation pipeline assume a linear coordinate space. Migrating a clinical pipeline to a pangenome reference is a software engineering project measured in person-years, and it is the reason adoption has trailed publication by so long.

“Migrating a clinical pipeline from a linear reference to a pangenome graph is a software project measured in person-years.”

An atlas, and a foundation model

On 20 November 2024 the Human Cell Atlas consortium published a Nature collection, 'The Human Cell Atlas: towards a first draft atlas', spanning 18 biological network atlases, together with a perspective on moving from a cell census toward a unified foundation model. The consortium, established in 2016, framed the milestone as a shift from data collection to integration.

That framing is the interesting part for computing professionals. An atlas assembled from thousands of independently generated single-cell datasets is a data-integration problem before it is a biology problem: batch effects, inconsistent ontologies and unequal donor representation determine what any model trained on it will learn.

From sequence to therapy

On 8 December 2023 the FDA approved Casgevy (exagamglogene autotemcel), developed by Vertex Pharmaceuticals and CRISPR Therapeutics — the first CRISPR-Cas9 gene-editing therapy approved in the United States, for sickle cell disease in patients aged 12 and over. The same day the agency approved Lyfgenia from bluebird bio, a lentiviral gene therapy with a different mechanism. Casgevy was reported at a list price of approximately $2.2 million per patient as a one-time treatment, addressing a US population of roughly 16,000 eligible patients aged 12 and over with severe disease.

Every step between a patient sample and an approved edit is now software-mediated: guide RNA design, off-target prediction, and the sequencing-based verification that the intended edit and no other occurred. Those tools sit inside a regulated therapeutic pathway, and the profession should treat their validation with the seriousness that implies.