Open-source single-cell pipelines get a modular alternative
Single-cell analysis has drifted toward end-to-end systems that promise speed and convenience, but often at the cost of flexibility. The new gcat library and its companion toolkit, its_a_scam, are built to push back against that trend by separating core genomic parsing from downstream single-cell processing and quality control.
The framing is straightforward: users should not have to surrender modularity simply to move from raw sequencing data to interpretable matrices. In practice, that means a researcher can keep using common open-source workflows for count matrix generation while still gaining access to alignment-level metrics that proprietary pipelines usually hide behind a fixed interface.
What gcat and its_a_scam are designed to do
At the base of the stack sits gcat, described as a C++ library for parsing genomic data and working across genomic coordinates. On top of it sits its_a_scam, which extends the same infrastructure into a broader set of tools for single-cell data processing and quality control.
The logic here is architectural rather than cosmetic. By splitting the low-level genomic engine from the single-cell workflow layer, the system can support multiple use cases without forcing every analysis through a single, rigid pipeline. That matters when a group wants to examine the data at different stages — before clustering, after annotation, or at the point where per-cell artifacts become biologically misleading.
One sentence captures the practical consequence: alignment files no longer have to remain a black box.
Per-cell quality control from alignment files
The most consequential feature is the ability to compute per-cell quality control directly from alignment files. Those metrics can then be used to assess each cell individually or to compare cell type differences after clustering or annotation. That expands quality control beyond simple count-based filtering and into a more granular view of how reads distribute across the genome.
In single-cell genomics, that extra resolution is not trivial. Cells with similar counts can still differ sharply in the regions they capture, the shape of their metagene profiles, or the breadth of genebody coverage. Those differences can reflect technical noise, but they can also reflect genuine biology, especially in heterogeneous samples where chromatin accessibility or transcriptional structure varies by lineage.
Its_a_scam is positioned to surface those patterns early. Rather than waiting until a downstream interpretation step exposes a problem, the toolkit gives users alignment-level summaries that can flag cell-specific irregularities or reveal cell state differences that would otherwise be flattened into a matrix.
Concordance with proprietary count generation
The preprint tests the approach in a peripheral blood mononuclear cell Multiome dataset. In that dataset, count matrices generated with the help of its_a_scam were highly concordant with matrices produced by proprietary pipelines.
That concordance is central to the argument. If an open-source workflow cannot reproduce the core matrix output that researchers expect from commercial systems, flexibility comes at too high a price. Here, the reported agreement suggests the modular route can preserve the standard output while opening the door to more transparent and customizable analysis.
The result also speaks to a broader issue in single-cell genomics: many labs want both reproducibility and control. Proprietary end-to-end systems can simplify onboarding, but they also constrain how data are handled after the first pass. A modular framework is more demanding operationally, yet it lets analysts inspect the underlying data structure rather than accepting a prepackaged summary as the final word.
Alignment metrics reveal cell type differences
Beyond matrix concordance, the alignment metrics exposed differences across cell types in aligned genomic regions, metagene profiles, and genebody coverage. Those signals matter because they show the toolkit is doing more than repackaging existing outputs. It is measuring features of read distribution that can separate one population from another even after the usual processing steps have already been completed.
In single-cell multiomic work, that kind of detail can help distinguish biological variation from technical degradation. A cell with unusual coverage across gene bodies may have a different transcriptional state, but it may also carry evidence of library preparation artifacts or mapping bias. A tool that quantifies those patterns at the level of individual cells gives analysts a sharper instrument for deciding which interpretation is more defensible.
The fact that these metrics diverge by cell type suggests the alignment layer is capturing biologically meaningful structure, not only quality-control noise. For users working with mixed populations, that is a meaningful extension of the standard pipeline.
Why the modular design matters for real workflows
The software is explicitly designed to integrate with commonly used open-source tools for scRNA-seq and scATAC-seq count matrix generation. That interoperability is one of its main strengths. Instead of asking users to abandon existing infrastructure, the toolkit slots into current pipelines and adds a layer of functionality that proprietary systems often keep bundled inside a closed workflow.
This modularity has a second benefit: it supports analysis beyond the default pipeline. Researchers can use the same framework for routine processing and for exploratory work that asks different questions of the same data. In a field where the same dataset may need to support quality control, annotation, integration, and cell-type comparison, that flexibility is not ornamental. It changes what can be asked without rebuilding the analysis from scratch.
A single pipeline can be convenient. A modular one can be interrogated.
Open-source availability and the limits of the preprint
Both tools are open source and available through GitHub, with separate documentation sites for gcat and its_a_scam. The manuscript also lists funding from the National Institutes of Health and the Rita Allen Foundation, and the authors declare no competing interest.
Because this is a preprint, the findings have not yet passed peer review. The report does not provide a clinical dataset in the medical sense, and it does not claim any patient-facing intervention. Its contribution is computational and methodological: a framework for more transparent single-cell processing, alignment-level quality control, and closer inspection of cell-specific genomic behavior.
The unresolved question is not whether the tools can generate matrices or compute per-cell metrics. It is how widely they will be adopted in labs already invested in proprietary workflows, and how well they will perform across datasets with different sequencing chemistries, tissue sources, and quality profiles. Those are the settings that will determine whether modular single-cell analysis remains a niche preference or becomes a standard alternative.