The team developed SNP2Cluster-Py as an open-source Python framework that intentionally separates genomic relatedness from epidemiological inference. The program accepts a precomputed pairwise Core SNP distance matrix alongside user-configured metadata. Users can analyse all eligible isolates together or stratify analysis by contexts such as sequence type (ST), geographic facility, epidemiological grouping, and ordered composite contexts. The workflow therefore treats genomic clustering as a reproducible, context-specific analytic step distinct from downstream epidemiological interpretation.
The software defines authoritative Core clusters within each context as connected components in an SNP-threshold graph. K-means assignments and silhouette widths are produced, but the study team retained those outputs only as exploratory diagnostics rather than primary cluster calls. The abstract did not report exact SNP-threshold values, nor did it specify default parameter settings for K-means or silhouette calculations; users will need to consult the code repository or full text for implementation defaults and tunable options.
Investigators reported a clear analytical separation between the generation of Core SNP clusters and the subsequent application of temporal or epidemiological criteria to infer possible transmission. Core clusters are determined strictly by graph connectivity at an SNP threshold. Analytical clustering via K-means and silhouette metrics is available but explicitly not used to override graph-based cluster assignments; those diagnostics are retained to inform exploratory analysis and sensitivity checks. The source did not provide granular methodological choices for threshold selection in the abstract, so practical threshold selection and justification are left to the user or to details in the repository.
The authors evaluated the workflow using 172 Klebsiella pneumoniae isolates collected from six neonatal-surveillance hospitals in South Africa. Using an ST-only context, the pipeline assigned 141 of these isolates to Core SNP clusters. When the context combined Facility + ST, assignment dropped to 134 isolates. These divergent counts illustrate that facility stratification can break genomic connections that span locations, removing cross-facility links that might otherwise be included in an ST-only analysis.
This behaviour makes the analytical choices explicit: genomic similarity does not automatically translate to epidemiologically relevant transmission. The study team used the dataset to demonstrate how changing context alters the set of isolates deemed part of authoritative Core clusters, rather than to assert confirmed transmission events.
A focused subset of 36 isolates from a single hospital, drawn from the same K. pneumoniae surveillance collection, was analysed separately. Both ST-only and the single-hospital context assigned 30 of 36 isolates to five Core clusters. When investigators applied Facility analysis together with temporal criteria, they retained 23 isolates distributed across four temporally compatible clusters. The comparison shows how adding facility and temporal constraints further refines cluster membership, yielding a smaller set of clusters consistent with contemporaneous sampling and local epidemiological context.
The authors also applied SNP2Cluster-Py to 168 Streptococcus pneumoniae carriage isolates from Ghana. ST-context analysis assigned 114 isolates to 38 Core clusters. Those Core clusters were then characterised across sampling locations and over time through a Community analysis. The pipeline thus supported stratified description of cluster composition by both genetic background (ST) and sampling context (community/location and time), enabling users to trace how Core clusters map to epidemiological metadata.
SNP2Cluster-Py v1.1.2 is distributed under the Apache License 2.0. The authors provided code and data links in the preprint, including GitHub and Zenodo entries for the software and a Bioproject accession for sequence data. The preprint states the workflow is pathogen-agnostic, intended for reuse across different surveillance contexts, and designed to produce reproducible Core SNP clusters that can be further filtered or interrogated with user-defined epidemiological and temporal criteria.
Users seeking implementation details should consult the linked repository and supplementary materials. The abstract does not specify runtime performance metrics, computational requirements, or recommended SNP thresholds; those technical specifics are likely in the full text or code documentation.
The authors noted that genomic similarity alone is insufficient to establish epidemiologically meaningful transmission; the framework intentionally maintains that distinction. The manuscript is a preprint and has not been certified by peer review. The abstract did not report exact SNP-threshold values, default parameter choices for K-means or silhouette analysis, or detailed computational performance benchmarks; those items are not described in the abstract and require consultation of the repository or full paper. Open questions therefore include guidance for threshold selection in different pathogens or surveillance settings, recommended metadata schemas for temporal compatibility checks, and validation against confirmed transmission chains—items not addressed in detail within the abstract.
For reproducibility the authors provided links to GitHub, Zenodo and a Bioproject accession; practitioners should review those resources when adopting the workflow and when deciding context definitions and SNP thresholds appropriate for their surveillance questions.