Finding. In human hematopoietic cells that already carry lentiviral barcodes, somatic mitochondrial DNA mutations below 10% per-cell heteroplasmy are not leftover noise. After molecule-level error filtering they make up about half of the lineage-informative calls. Keep the full heteroplasmy spectrum and clonal assignment's area under the precision-recall curve roughly doubles versus a greater-than-10% cutoff. Single-molecule-supported calls help when you keep them as a class.
Why this paper matters
Mitochondrial DNA is the endogenous barcode everyone already has. It mutates, it is present in many copies, and it can be read next to cell state. The field responded to the obvious error problem by keeping only high-heteroplasmy variants: easier to call, fewer of them, and more likely to have been through selection. That gate buys precision and spends resolution.
Weng, Sankaran, Weissman and colleagues put an independent clone label under the calls. Lentiviral barcodes in human hematopoietic cells are the ground truth. Once molecule-level filters are applied, the low-heteroplasmy half of the informative set is real lineage. That is a methods paper with a corpus-level consequence. Any atlas, tumor phylogeny, or aging clone map that discarded variants below 10% heteroplasmy was not being conservative. It was throwing away about half of its mitochondrial DNA clone marks, at least in this blood-cell system.
The paper does not claim that low-heteroplasmy mutations are selectively inert, or that every singleton is a barcode. It claims they encode bona fide clonal history when error is handled at the molecule, and that keeping them collectively improves assignment. That is enough to change a default. It is not enough to reopen every published clone tree without reading how those authors filtered.
What they actually measured
The design is a concordance test. Human hematopoietic cells receive lentiviral barcodes. Mitochondrial DNA mutations are called across the heteroplasmy range. Stringent molecule-level error filtering is applied. Clonal assignment is scored against the viral barcodes with and without a greater-than-10% heteroplasmy cutoff, and again with single-molecule-supported calls kept or dropped as a class. The summary statistics in the abstract are the ones that matter: about half of lineage-informative calls sit below 10%; full-spectrum calling approximately doubles area under the precision-recall curve; single-molecule calls help in aggregate.
Missing from the abstract: chemistry (amplicon, long read, incidental RNA-seq mitochondrial reads), cell numbers, barcode complexity, the exact molecule-level filters, and whether the cells were cultured, differentiated, or transplanted. Those details decide whether you can port the result to a tumor biopsy or a brain snRNA-seq object. The hematopoietic system is a favorable test bed (dividing cells, established barcoding). Favorable is not universal.
How to read the score
Low eighties. Direct mitochondrial DNA methods, an independent ground truth, a quantitative gain, and a default the field can actually change. Confidence is high for the claim as scoped: barcoded human hematopoietic cells, after strict molecule-level filters. Confidence is not high that every tissue and every library type will see a twofold lift, or that low-heteroplasmy variants are free of selection. This is not a disease paper and not a dysfunction paper. It is a resolution paper. The score reflects that: important for anyone who uses the organelle as a clock or a clone stamp; not a therapeutic hit.
What to do with it
If your pipeline still hard-gates mitochondrial DNA variants at 10% heteroplasmy, treat that gate as a hypothesis, not a standard. Re-run assignment with the tail in, after molecule-level filters, and report both curves. When you ingest lineage papers, tag whether they discarded the low-heteroplasmy class. Pull the filter definition and the precision-recall supplement before quoting the doubling. Do not advertise "twice the clones in every human tissue." Do not treat single-molecule calls as individually sacred; the abstract's claim is collective recovery.