Two locked operators report mixed-spot TME composition on CosMx carcinomas

This study presents two locked operators around a training-free composition baseline in a standalone SPGD-TME Methods implementation so that, for the locked or explicitly eligible malignant–neighbor comparison, a per-spot malignant fraction is written only where the reference can separate the malignant program from its neighbor: a KEEP/ABSTAIN reportability gate that withholds the malignant coordinate once the malignant–neighbor cosine reaches a frozen cutoff c ⋆ = 0.80 and a Patient_ID donor split that builds mixed spots and reference from disjoint patients. On four-patient CosMx basal-cell carcinoma, the operators recover named tissue objects: ulcerated-nodular Patient B is tumor-rich relative to Patient A, the wound axis on D-from-C decreases tumor fraction from 0.8425 to 0.4665, while fibroblast and MoMacDC increase and directed transfer recovers composition most tightly on B from A (RMSE 0.0659, spot-level PCC 0.9528). The frozen cosine reports KEEP and, therefore, a tumor fraction, on CosMx BCC (cosine 0.6037), independent CosMx NSCLC (tumor versus fibroblast 0.6482; directed-transfer RMSE 0.1066, spot-level PCC 0.8415), CosMx colorectal carcinoma (0.5058), and CosMx HCC (0.6486). It returns ABSTAIN, holding the malignant coordinate missing and continuing the remaining types on the simplex, CosMx PDAC (0.8616), HNSCC openST (0.9727), breast Xenium (0.9802), and an orthogonal Xenium FLEX reference (0.9770). CosMx PDAC abstains on the same platform that carries all four KEEP calls, so the observed pattern is consistent with collinear type pairs; the present library-level design does not establish platform independence. A repeated 400-spot openST build takes 16.103 ± 2.891 s (five warm runs; CV 17.95 % ), of which the in-memory reportability decision is 0.001 s so that reportability remains a small operation relative to the fixed baseline fit.

Authors

Institutions

Publication Details

Journal
Frontiers in Genetics
Published
2026-11-18
DOI
https://doi.org/10.3389/fgene.2026.1981011
Primary Topic
Single-cell and spatial transcriptomics
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Two locked operators report mixed-spot TME composition on CosMx carcinomas

Tianfei Ran, Zeyu Fu, Lei Ran, Rui Yang
3 links
Frontiers in Genetics
Single-cell and spatial transcriptomics
article

Two locked operators report mixed-spot TME composition on CosMx carcinomas

Tianfei Ran, Zeyu Fu, Lei Ran, Rui Yang
article en

Abstract

This study presents two locked operators around a training-free composition baseline in a standalone SPGD-TME Methods implementation so that, for the locked or explicitly eligible malignant–neighbor comparison, a per-spot malignant fraction is written only where the reference can separate the malignant program from its neighbor: a KEEP/ABSTAIN reportability gate that withholds the malignant coordinate once the malignant–neighbor cosine reaches a frozen cutoff c ⋆ = 0.80 and a Patient_ID donor split that builds mixed spots and reference from disjoint patients. On four-patient CosMx basal-cell carcinoma, the operators recover named tissue objects: ulcerated-nodular Patient B is tumor-rich relative to Patient A, the wound axis on D-from-C decreases tumor fraction from 0.8425 to 0.4665, while fibroblast and MoMacDC increase and directed transfer recovers composition most tightly on B from A (RMSE 0.0659, spot-level PCC 0.9528). The frozen cosine reports KEEP and, therefore, a tumor fraction, on CosMx BCC (cosine 0.6037), independent CosMx NSCLC (tumor versus fibroblast 0.6482; directed-transfer RMSE 0.1066, spot-level PCC 0.8415), CosMx colorectal carcinoma (0.5058), and CosMx HCC (0.6486). It returns ABSTAIN, holding the malignant coordinate missing and continuing the remaining types on the simplex, CosMx PDAC (0.8616), HNSCC openST (0.9727), breast Xenium (0.9802), and an orthogonal Xenium FLEX reference (0.9770). CosMx PDAC abstains on the same platform that carries all four KEEP calls, so the observed pattern is consistent with collinear type pairs; the present library-level design does not establish platform independence. A repeated 400-spot openST build takes 16.103 ± 2.891 s (five warm runs; CV 17.95 % ), of which the in-memory reportability decision is 0.001 s so that reportability remains a small operation relative to the fixed baseline fit.

Frontiers in GeneticsVol. 17
Army Medical University (CN), Xinqiao Hospital (CN), Southwest Hospital (CN)
Ministry of Science and Technology of the People's Republic of China, National Key Research and Development Program of China
Good health and well-being
Single-cell and spatial transcriptomics
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon