Barrera-Suarez, M. A.; Gupta, V. K.; Zhao, X.; Koller, A. L.; Zhao, C. Y.; Suh, G. A.; Karnatovskaia, L. V.; O'Brien, E. K.; Kronzer, V. L.; Sung, J.

Background: Upper-airway and respiratory conditions impose a large and growing global burden, yet there is no standardized way to assess whether a nasal microbiome is healthy. Currently, monitoring remains reactive, beginning only after symptoms manifest. The nasal cavity is well suited to proactive monitoring: it shapes respiratory health and pathogen colonization resistance, and can be sampled non-invasively and repeatedly. To address this gap, we introduce the Nasal Microbiome Wellness Index (NMWI), a disease-agnostic, continuous score of nasal microbiome health derived from LASSO-penalized logistic regression. Rather than counting taxa, it learns which taxa (and in what balance) characterize a healthy nose and returns the predicted log-odds that a profile resembles a healthy state. The index was trained on 1654 nasal 16S rRNA gene amplicon sequencing samples (589 healthy, 1065 non-healthy) pooled from 27 publicly available studies, uniformly reprocessed through a single computational pipeline. Results: The NMWI comprises an interpretable signature of 24 taxa whose combined relative abundances determine the health-associated log-odds. Health-associated genera such as Corynebacterium and Cutibacterium raised the score, while dysbiosis-associated genera such as Pseudomonas and Escherichia-Shigella lowered it. The NMWI substantially outperformed the Shannon, Simpson, and Chao1 diversity indices, which showed negligible, directionally inconsistent separation between healthy and non-healthy samples (|Cliff's delta| = 0.01-0.20), whereas the NMWI produced large, consistent separation (delta = 0.65). It achieved a balanced accuracy of 74.37% on the training data (resubstitution estimate), and 73.43% under repeated 10-times 10-fold cross-validation. Performance remained stable at mean balanced accuracy of 73.82% under a leave-one-study-out framework, reflecting cross-study generalizability. In independent external cohorts, the balanced accuracy was 71.49%, and leave-one-disease-out analysis (in which each disease condition was withheld from training) showed a mean balanced accuracy of 63.92% across unseen conditions, consistent with a disease-agnostic design. The index also generalized across heterogeneous datasets spanning multiple 16S rRNA gene hypervariable regions; to our knowledge, this is the first demonstration of such cross-study, cross-region transferability for the nasal cavity. Conclusions: The NMWI distills a complex nasal microbial profile into a single interpretable score computed directly from the 16S rRNA gene data that dominate existing nasal research, making it immediately applicable to published and future datasets without re-sequencing. By replacing descriptive, diversity-based comparison with a quantitative standard, it offers a reproducible, open-source foundation for cross-study benchmarking, individual-level phenotyping, and longitudinal respiratory wellness monitoring.