2026-10-15T03:01:07Z https://www.scidb.cn/oai

10.57760/sciencedb.0134y 2026-09-21T11:21:10Z

2026-09-21 Analysis code for: A peripheral-blood eight-gene transcriptomic signature for tuberculosis with relevance to extra-pulmonary disease doi:10.57760/sciencedb.0134y en This dataset contains the complete analysis code for the manuscript "A peripheral-blood eight-gene transcriptomic signature for tuberculosis with relevance to extra-pulmonary disease: multicohort external validation, Mendelian randomization, and mechanistic insights", deposited to support full reproducibility of the reported results.Contents. Ten analysis modules comprising 90 scripts (Python and R): (01) discovery-cohort differential expression, LASSO derivation of the eight-gene diagnostic signature, and bootstrap stability assessment; (02) transfer-based external validation across five independent cohorts with fixed model coefficients and no re-estimation, head-to-head benchmarking against RISK6, paired DeLong testing, and decision-curve analysis; (03) two-sample Mendelian randomization (CRP to tuberculosis, 138 instruments, including IVW, MR-Egger, weighted median, weighted mode, Cochran's Q, Egger intercept, and leave-one-out sensitivity analysis); (04) immune-cell deconvolution; (05) pan-programmed-cell-death scoring and residualization against a 12-cell-type immune-infiltration matrix; (06) cross-tissue bridging of the blood signature to spinal-lesion and bone-marrow mesenchymal stromal cell transcriptomes; (07) exploratory plasma miRNA analysis; (08) functional enrichment (GO/Reactome/KEGG); (09) shared utility scripts; and (10) scripts generating all main and supplementary figures.Data sources. All source data are publicly available and were obtained from NCBI GEO (GSE89403, GSE83456, GSE19491, GSE37250, GSE144127, GSE185372, GSE268366, GSE222013, GSE62525, GSE70476, GSE225679) and OpenGWAS (e.g. ieu-b-35). This deposit redistributes no third-party raw or expression data.Processing. All analyses were performed in Python 3.13.12, primarily relying on numpy 2.4.6, pandas 2.3.3, scikit-learn 1.9.0, scipy 1.18.0, statsmodels 0.14.6, matplotlib 3.11.0, seaborn 0.13.2, shap 0.52.0, requests 2.34.2 and primer3-py 2.3.0 (a fully pinned list is provided in requirements.txt). Random seeds are fixed (random_state=2026 for modelling; random_state=0 for discovery-cohort cross-validation splits).Quality control. All scripts were re-run end to end in modules 01 through 10 prior to deposition, and the outputs match the values reported in the manuscript. A known limitation, namely that the internal discovery-cohort cross-validation did not use participant-level grouping and may overestimate internal AUC, whereas the five external cohorts are fully independent and unaffected, is stated explicitly in README.md.Licence and citation. Released under the MIT licence. Please cite this record using its assigned DOI, together with the manuscript above. tuberculosis; transcriptomic signature; LASSO; external validation; Mendelian randomization; biomarker; spinal tuberculosis; host immune response; bioinformatics pipeline Zixing Xu Chuanrong Chen Changyi Jiang Xinhao Huang Zixuan Lin Weihong Xu PUBLIC https://mit-license.org dataset Science Data Bank