RIT-Bridge v0.2

Radioimmune transcriptomic scoring workbench for bulk RNA-seq cohorts.

Model 0.2-alpha
Research use only Raw matrix processed locally

Workbench

Accepted: CSV, TSV, TXT, or gzip-compressed text in modern browsers. First column should contain gene symbols or sample IDs.
1.Normalize 2.Score 3.Explain 4.Export
Choose a matrix to begin.

Analysis Results

Cohort-relative research scores
Dataset: No dataset Samples: 0 Model: 0.2-alpha Processing: Local browser
Samples 0
Median RIT-Gain --
High Candidates 0
Mean Coverage --
Run the model to render the score map.
Download

My Analysis History

Sign in to save and revisit derived RIT-Bridge results.

runs
Derived results only. Raw matrices are not stored.
Your raw expression matrix remains in this browser and is not stored in cloud history.
Research documentation Method, evidence, validation, input guidance, limitations, and release notes

Result Interpretation

RIT-Gain v0.2

Higher values mean the sample resembles a radio-immunotherapy gain-like state in this dataset. Interpret it as a research prioritization score, not a predicted treatment benefit.

RT convertibility v0.2

Higher values indicate stronger RT-priming potential with immune competence and lower barrier burden. It is useful for identifying samples that may warrant deeper radioimmune analysis.

Uncertainty

Uncertainty increases when evidence alignment is weak or core modules disagree. Low confidence should trigger manual review rather than exclusion.

Method

RIT-Bridge v0.2-alpha uses curated radioimmune signatures, within-dataset standardization, evidence-weighted components, and an explicit uncertainty layer.

RIT-Gain = 0.28 RT_priming + 0.16 ICI_inflamed + 0.30 immune_effector - 0.28 immune_barrier + 0.06 checkpoint_context + 0.04 proliferation_context
RT convertibility = 0.48 RT_priming + 0.24 immune_effector - 0.28 immune_barrier - 0.12 positive(ICI_inflamed)
  • RT priming combines STING/DNA-damage and type I IFN programs.
  • Immune effector combines cytotoxic T-cell and NK activity programs.
  • Immune barrier combines CAF/TGF-beta, myeloid suppression, hypoxia, and EMT programs.
  • Scores are standardized within the uploaded matrix, so interpretation is cohort-relative.

Evidence Atlas

GSE253564 + GSE248378 supportive discovery
NSCLC durvalumab with or without stereotactic radiation; directional discovery layer, not prospective validation.
GSE190905 mechanistic support
NSCLC SABR/SBRT immune-transition and TCR support; peripheral immune evidence, not tumor-site prediction.
GSE329813 + GSE292098 indirect spatial context
NSCLC immunotherapy spatial ecosystem interpretation; useful for effector/barrier biology but lacks radiotherapy exposure.
GSE295599 + GSE277984 mixed public stress tests
Open RT+ICI and NSCLC SBRT tests show mixed or negative behavior, which narrows the claim and keeps the model conservative.
PEMBRO-RT / EGAS50000000277 pending access
Best direct NSCLC SBRT+pembrolizumab validation candidate, but not current evidence until EGA access and analysis are complete.

Validation and QA

Robustness

Weight sensitivity analysis preserved stable v0.2-alpha ranking under perturbation, with worst 5th-percentile rank correlation of 0.947. RT priming produced the largest ablation shift.

Public stress tests

GSE295599 provides an open RT+ICI tumor stress test with mixed results; GSE277984 anchors lung SBRT biology through public NSCLC PBMC data. These tests narrow the claim rather than proving response prediction.

Implementation QA

Python and browser scoring match on the demo matrix, static public-release QA has 0 hard failures, and Phase 15 visual-risk/copy checks have 0 hard failures.

Future validation

PEMBRO-RT remains pending access and is the preferred direct NSCLC SBRT+pembrolizumab validation candidate if controlled-access EGA approval and download succeed.

Public Readiness

PEMBRO-RT EGA access is no longer required for the first public website. It remains a future validation upgrade.

Allowed

Research-use, hypothesis-generating radioimmune biology workbench with transparent evidence boundaries.

Not allowed

No patient-level treatment recommendation, no clinical response prediction, and no PEMBRO-RT validation claim yet.

  • The public-facing navigation is organized around analysis, saved runs, evidence and validation, and consolidated documentation.
  • Public access enabled. Public access was enabled after explicit owner approval; the website remains research-use only.

Version Notes

v0.2-alpha scoring model

Frozen default score for the website. It uses evidence-weighted radioimmune signatures and remains the only default browser score. v0.2-alpha remains a hypothesis-generating research model, not a validated clinical predictor.

Adaptive v0.3-candidate

Experimental constrained layer for context-specific emphasis. It is documented but not deployed as the default model.

Public validation record

Phase 11-12 public datasets are retained as transparent stress tests and lung radiotherapy anchors, not clinical validation claims.

Release QA record

Phase 14-15 checks cover static release QA, reproducibility indexing, layout risk, and public claim wording. Browser pixel screenshots remain optional if automation is available.

Input Templates

Genes in rows

First column is gene symbol. Remaining columns are samples. This is the recommended format for bulk RNA-seq matrices.

Genes in columns

First column is sample ID. Remaining columns are gene symbols. Use this when your matrix is already sample by gene.

For large matrices or compressed GEO-style files, the Python CLI remains the most reliable path. The browser workbench is intended for simple matrices and quick review.

Limitations

  • RIT-Bridge v0.2-alpha is not clinically validated and must not be used for treatment selection.
  • Direct RT+ICI transcriptomic cohorts are small, so effect sizes should be treated as directional evidence.
  • Single-cell/TCR support comes from immune-transition evidence, not definitive tumor-site response validation.
  • Spatial datasets support effector and barrier interpretation, but they are not complete direct RT+ICI validation cohorts.
  • Uploaded scores are relative to the submitted matrix; cross-study comparison requires controlled preprocessing.

About the project

RIT-Bridge

Developed and maintained by Yueyang Yu.

Contributions: project conception, radioimmune framework design, evidence curation, model development, implementation, validation strategy, and public release stewardship.

Suggested citation

Yu Y. RIT-Bridge v0.2: a research-use framework for estimating radiotherapy-induced immunotherapy gain from transcriptomic data. 2026.

RIT-Bridge Account

Scoring stays in your browser. Cloud history receives derived scores and coverage summaries only after you choose Save to My History. Use pseudonymous sample IDs and dataset labels; never enter patient identifiers.