Research

Real methods. An open benchmark. Honest results.

We'd rather show the evidence than assert the vision. Two pieces of work anchor Span: a transparent detection method and a public benchmark whose central finding is a negative result we report in full.

Methods paper

Seeing below the limit of detection.

The Span detector, a censored-Poisson Bayesian latent-growth (CP-BLG) change-point test, reads the rate of ctDNA detections over time to flag resistance before the variant is callable on any single blood test. It has no trainable parameters; the advantage is structural.

Read on arXiv PDF Code
Span's lead time over a commodity snapshot across disease regimes, at a matched false-alarm rate.
Lead time across disease regimes at a matched 10% false-alarm rate (synthetic evaluation, 5 seeds). The gap is largest in the clinically hard, indolent regime.
Open benchmark

OncoTraj: an honest floor for what snapshots can predict.

OncoTraj harmonizes real-world clinical-genomic data into one schema with frozen, leakage-audited splits, and ships reproducible baselines. Its purpose is to establish a public floor, and an honest ceiling, for longitudinal resistance prediction in EGFR-mutant NSCLC on first-line osimertinib.

Read on arXiv Benchmark & code
SourcePatientsAccess
MSK-CHORD672cBioPortal (open)
FLAURA molecular-resistance supplement107Public supplements
AACR Project GENIE BPC (NSCLC)34Registered access
v1 total813Patient-level test split n=122
Headline finding

On clean within-source evaluation, snapshot-feature models do not beat chance on any of the three tasks at this cohort size.

The binding constraint is the data modality, single-timepoint tissue/ctDNA snapshots, not the model. Serial ctDNA trajectories, not a better algorithm, are the precondition for above-chance performance. That negative result is the contribution: it tells the field where the signal must come from.

Example: the 12-month landmark progression task (within MSK-CHORD only, so dataset membership can't leak into the label) reaches a logistic AUC of just 0.596 [0.478, 0.714], a confidence interval that crosses 0.50.

Results computed with bootstrap 95% confidence intervals (oncotraj-eval schema 1.1.0). We do not claim above-chance within-source performance where the data does not support it.

Why this de-risks the thesis

The negative result is the reason Span exists.

If snapshots could predict resistance, you wouldn't need us. OncoTraj shows, rigorously, on real patients, that they can't. The signal lives in the trajectory, which is exactly what the Span detector is built to read.

See the method Talk to us
Publications

Preprints.

The full write-ups behind the method and the benchmark, open to read, run, and check.

Method · Preprint

Seeing below the limit of detection: a censored-Poisson Bayesian latent-growth change-point detector for serial ctDNA in HR+/HER2− metastatic breast cancer

Aarchi Singh Thakur, Abhijoy Sarkar  ·  arXiv:2606.11876  ·  June 2026

arXiv PDF Code
Benchmark · Preprint

OncoTraj: a public benchmark for longitudinal resistance prediction in EGFR-mutant non-small-cell lung cancer on osimertinib

Abhijoy Sarkar, Aarchi Singh Thakur  ·  arXiv:2606.11144  ·  June 2026

arXiv PDF Code