AmbioSpire P003/T001: publication source data and reproducibility package (version 1.0.0)

This dataset and reproducibility package supports AmbioSpire P003/T001, a prospectively frozen evaluation of somatic small-variant calling procedures across benchmark contexts. The analytical procedure was developed using HCC1395/HCC1395BL and SEQC2 resources and subsequently evaluated, without retuning, against the protected NIST/GIAB HG008 v0.3 somatic benchmark. The study compares Mutect2, Strelka2 and DeepSomatic together with deterministic union, majority and strict-intersection ensembles using normalized exact CHROM/POS/REF/ALT allele identity. The primary evaluation region was the HG008 v0.3 nogermlineinterference region; the all region was used as a prespecified sensitivity analysis. The archive contains release-safe derived aggregate results, publication tables, figure-supporting data, five main figures, four Extended Data items, software and benchmark-resource metadata, provenance summaries, reproducibility documentation and SHA-256 manifests. It also includes prespecified secondary analyses of caller-support patterns, INDEL length, germline context, a within-50-bp exploratory region and all-region bootstrap uncertainty. These secondary analyses were specified before HG008 access and executed after completion of the primary external evaluation. Protected benchmark files, reference sequences, FASTQ/BAM/CRAM data, large caller VCFs, third-party software and binaries, private provenance records, manuscript files and the historical executable scoring/bootstrap source are not redistributed. An independent internal reproducibility review reproduced the retained numerical results. The work is research-only and does not establish clinical validity or population-level generalization.