Pattern Computer and Medical College of Wisconsin Report Breakthrough Point-of-Care Saliva Cancer Screening Results
核心洞察
Pattern Computer (搜索) and the Medical College of Wisconsin published initial results for a rapid, point-of-care tier-1 cancer (搜索) screening test using saliva transmission hyperspectral spectroscopy paired with machine learning.
The ProSpectral device scanned just two drops of saliva in approximately 3 seconds, supporting sub-minute time-to-results with no reagent costs, versus a two-week turnaround for sequencing-based cfDNA blood assays.
Across a 251-participant cohort, the model achieved 91% specificity, 44% sensitivity, 61% balanced accuracy, and an F1-score of 0.55, with statistical saturation analysis suggesting specificity above 98% is attainable with fewer than 500 samples.
Pattern Computer (搜索), Inc., in collaboration with the Medical College of Wisconsin (MCW), the Froedtert & the Medical College of Wisconsin Clinical Cancer (搜索) Center, and the Linda T. and John A. Mellowes Center for Genomic Sciences and Precision Medicine at MCW, has announced the publication of initial results from research on rapid, point-of-care (PoC) tier-1 cancer screening using Pattern's ProSpectral device. The findings, which pair saliva transmission hyperspectral spectroscopy with advanced machine learning, were presented at the annual meeting of the American Society of Clinical Oncology (ASCO), published in the Journal of Clinical Oncology, and showcased as a poster at the annual meeting of the Association for Diagnostics & Laboratory Medicine (ADLM) held July 26–30, 2026 in Anaheim, CA.
The study is led by Razelle Kurzrock, MD, FACP, Director of the Center for Precision Oncology and Rare Cancers at the MCW Cancer (搜索) Center and Associate Director of Precision Oncology at the Mellowes Center for Genomic Sciences and Precision Medicine. Dr. Kurzrock, recognized by GPS Scholar as number 1 in the precision medicine field worldwide, is joined by co-leaders Hui Zi Chen, MD, PhD, and Ann Maguire, MD, MPH, at the Froedtert & MCW Clinical Cancer Center.
Solving the Tier-1 Screening Bottleneck
Effective tier-1 cancer (搜索) screening can benefit from rapid, inexpensive, and highly specific testing to minimize unnecessary downstream workups and patient anxiety. Current centralized laboratory workflows, such as sequencing-based cfDNA blood assays, typically require a two-week turnaround time and high reagent costs. By contrast, the newly evaluated method can support sub-minute time-to-results at the point of care, completely avoid reagent costs, and discriminate as well as traditional sequencing-based cfDNA methylation blood assays at a comparable cohort scale.
Study Design and Methodology
The clinical evaluation was conducted under an IRB-approved protocol (PREDICT NCT05802069) using saliva collected from outpatient consented participants at MCW facilities. The cohort comprised 251 total participants, including 99 from a lung cancer (搜索) clinic and 152 from a high-risk clinic. Just two drops of saliva were scanned in approximately 3 seconds using the ProSpectral hyperspectral spectrophotometric device to capture transmission spectra.
The data were analyzed using Pattern Computer (搜索)'s proprietary Pattern Discovery Engine (PDE). PDE models generate human-readable symbolic equations in the spectral domain, providing true explainability and allowing researchers to calibrate operating thresholds specifically for high-specificity, low false-positive tier-1 operations.
Promising High-Specificity Results
Patient data were categorized into four distinct clinical states: cancer (搜索)-positive (n = 99), hereditary cancer-predisposition without diagnosis (n = 86), no evidence of disease (NED) while on treatment (n = 16), and NED at least 2 months post-treatment (n = 50). The predictive models successfully maintained robust discrimination under strict settings designed to eliminate false positives.
Initial performance metrics across the 251-sample cohort demonstrated a specificity of 91%, a balanced accuracy of 61%, a sensitivity of 44%, and an F1-score of 0.55. Statistical saturation analysis indicates that specificity exceeding 98% is likely attainable with a cohort of fewer than 500 samples, establishing a clear path toward a clinically actionable screening protocol.
Next Steps and Ongoing Work
"The ProSpectral data demonstrates that saliva transmission hyperspectral spectroscopy, powered by our Pattern Discovery Engine, establishes a highly scalable, near-real-time paradigm for early cancer (搜索) screening and triage," commented Mark R. Anderson, Pattern Chair and CEO. "We are not merely adjusting current diagnostic workflows; we are completely disrupting them. This innovation shifts the diagnostic timeline from weeks to a sub-minute at the point of care, equaling traditional sequencing-based blood assays at scale, while completely removing chemical reagent costs from the equation."
Anderson added that ongoing development is focused on attributing specific discriminative spectral signatures to underlying biological markers. The joint team plans to use orthogonal assays, including fractionation and targeted mass spectrometry, to identify the contributing analytes and host-response biomarkers, which will further improve the system's interpretability ahead of upcoming prospective validation trials.