Somatic mutations arise across life and their functional impact depends on the cell populations in which they occur. The authors developed SniffCell, a pipeline that exploits DNA methylation information inherent to native long-read sequencing molecules to map somatic variant-supporting reads back to methylation-resolvable cell types. The stated goals were to enable genome-wide somatic variant discovery from bulk long-read whole-genome sequencing (WGS) while preserving cell-type context, and to validate cell-type-specific variant calls in both blood and brain tissues.
SniffCell first builds methylation signatures that discriminate cell types across eight tissues using methylation patterns carried on native long reads. These signatures are derived from long-read WGS molecules rather than relying on external single-cell methylomes, allowing the method to operate directly on bulk native-read data. The approach emphasizes methylation sites that resolve cell identity and uses those sites to define signatures suitable for assigning individual long reads to cell types.
After establishing discriminatory methylation patterns, SniffCell assigns individual long reads to cell types and then performs cell-type-specific variant calling on reads supporting somatic variants. The authors benchmarked the approach using peripheral blood mononuclear cells and multiple brain datasets. Validation included recovering identities of sorted cell populations and confirming assignments with purified fractions from immune cells, neurons, and oligodendrocytes. The source reports that SniffCell successfully recovered sorted cell identities and validated cell-type-specific variant assignments in these experimental settings.
Applied to blood, SniffCell recovered lineage-restricted antigen receptor rearrangements, illustrating the method's ability to localize highly cell-type-specific alleles. Additionally, the pipeline localized a somatic tandem-repeat expansion to T cells, demonstrating that complex alleles such as tandem-repeat expansions can be assigned to specific immune lineages using methylation-resolved long reads.
In frontal cortex samples, SniffCell identified recurrent neuron-specific tandem-repeat expansions in multiple genes, including FGF14, LRRC7 and SH3RF3. These expansions were observed specifically on reads assigned to neuronal methylation signatures, indicating cell-type-restricted somatic instability. The findings highlight recurrent neuron-associated repeat instability in human brain tissue detectable from bulk long-read data when reads are resolved by cell-type methylation patterns.
Across three brain cohorts that together included 172 donors, recurrent neuron-associated expansions were enriched for GAA-rich motifs. Where donors had matched blood samples, neuron-associated alleles diverged more strongly from the inherited repeat length than oligodendrocyte-associated alleles, which more often tracked the inherited repeat length. These cohort-scale observations suggest a cell-type difference in somatic repeat instability, with neurons showing greater divergence from germline repeat length compared with oligodendrocytes in the datasets analyzed.
The authors report that SniffCell is available at the project GitHub repository (https://github.com/Fu-Yilei/SniffCell) and analysis scripts are available at https://github.com/Fu-Yilei/SniffCell-analysis. An accompanying tool, mTRplotter, is available at https://github.com/Fu-Yilei/mTRplotter. Data from four QSBB brains (bulk and sorted long-read WGS, short-read scWGS) will be released for controlled access via the ASAP CRN Cloud and are to be described in Zenodo. SMAHT and CARD cohort data accessions and repositories are noted in the source; specific controlled-access locations were provided there.
SniffCell demonstrates that native long-read WGS can be transformed into a cell-type-aware resource for somatic variant discovery by exploiting endogenous methylation signals on sequencing molecules. This enables cell-type-specific localization of complex somatic alleles, including tandem-repeat expansions and antigen receptor rearrangements, from bulk samples without single-cell isolation.
Limitations and details not reported in the source: the article abstract and accompanying information describe validation and cohort findings but do not present specific sensitivity, false-discovery rates, or per-variant performance metrics in the text excerpt provided here. The abstract notes benchmarking and validation in blood and brain but does not report numerical performance measures or exhaustive methodological parameters in the supplied source excerpt. Any further performance statistics, parameter choices, or complete methodological steps should be obtained directly from the full preprint, associated supplementary materials, or the public code repositories cited by the authors.
By using methylation carried on native long reads, SniffCell assigns somatic variant-supporting molecules to cell types and enables cell-type-specific variant calling from bulk long-read WGS. The method recovered expected cell identities in validation experiments, localized complex alleles such as antigen receptor rearrangements and tandem-repeat expansions to specific cell types, and revealed recurrent neuron-associated repeat instability enriched for GAA-rich motifs across multiple brain cohorts. The code and tools are publicly available and several datasets are slated for controlled-access release as described by the authors.