STAT
STAT / Experiments / Single-query benchmark

Single-query benchmark

Forty analysis requests on three datasets: human DLPFC (Visium), mouse brain (MERFISH) and human breast cancer (Xenium). Thirty cover eleven analysis categories, from cell-type annotation and spatial domains to cell-cell communication and deconvolution; four are multi-step requests and six are boundary cases (one infeasible, three out of scope, two with a missing prerequisite).

STAT answered every request four times: Run 1 with Claude Sonnet 4.6, Runs 2–4 with DeepSeek V4.1 Flash at maximum reasoning effort (seeds 10, 20, 30). When STAT asked a question, a scripted user answered from the request's answer bank. Each answer was rated 1–5 by a Claude Opus 5.5 rater following a fixed rubric and per-request criteria; a rating of 3 or more counts as a success. The benchmark compared ten systems; this page lists STAT's runs.

160 runs
ConditionRunRating
Q01Annotate cell types on this tissue. It is a mouse brain cortex sample.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q02Identify cell types in this breast cancer tissue sample. I have a single-cell reference dataset at './breast_cancer_reference.h5ad' with cell type annotations that you can optionally use.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q03What cell types are present in this tissue?
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q04Can you identify the different cell populations here? I have a reference scRNA-seq dataset at './breast_cancer_reference.h5ad' if that helps.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q05Identify spatial domains in this brain tissue. There should be about 7 distinct regions including cortical layers and white matter.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q06Can you find the spatial structure in this tissue?
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 4/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 4/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 4/5 · successfulReplay →
Q07Identify distinct tissue regions based on spatial gene expression.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q08Find genes that show significant spatial expression patterns across this tissue.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q09Which genes have interesting spatial patterns?
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q10Identify spatially variable genes in this mouse brain section.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q11Analyze ligand-receptor interactions between cell types in this mouse brain tissue.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q12Find cell-cell communication patterns between tumor and immune cells in this human breast cancer sample.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q13How are the cells communicating with each other?
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q14Test whether certain cell types tend to co-localize or avoid each other spatially in this mouse brain tissue.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q15Analyze spatial co-localization patterns between cell types in this breast cancer sample.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q16Are certain cell types found near each other in this tissue?
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q17Estimate cell type proportions for each spot in this Visium tissue. I have a single-cell reference dataset at './dlpfc_reference.h5ad' with cell type annotations.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q18What cell types make up each spot in this tissue?
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q19Identify spatial microenvironments (niches) based on the local cell type composition in this mouse brain tissue.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 4/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q20Find the spatial neighborhoods in this tissue.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q21Find differentially expressed genes between the different spatial domains in this tissue.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 4/5 · successfulReplay →
Q22What genes distinguish the tumor regions from the surrounding stroma?
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q23Run gene ontology enrichment analysis on these marker genes: RELN, CUX2, PCP4, MBP, MOBP, NTNG2, TLE4. The species is human.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q24What biological pathways are active in the tumor microenvironment?
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 4/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 4/5 · successfulReplay →
Q25Integrate these four brain tissue slices, correcting for batch effects while preserving biological variation.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q26These two tissue sections are from the same brain region but look quite different. Can you make them comparable?
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q27Give me a summary of this dataset — how many spots, how many genes, what spatial extent, and what annotations are already present.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q28Show the top 10 most highly expressed genes across all cells in this tissue.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q29Give me a complete analysis of this spatial transcriptomics data.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q30What can you tell me about this tissue?
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 4/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q31First annotate cell types, then analyze which cell types tend to co-localize in this mouse brain tissue.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q32Identify spatial domains in this brain tissue (about 7 regions), then find genes that are differentially expressed between them.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q33Annotate cell types in this breast cancer sample using the reference at './breast_cancer_reference.h5ad', then analyze cell-cell communication between tumor and immune cells.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q34Identify cell types, find marker genes for each type, then run pathway enrichment on the tumor cell markers.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q35Estimate cell type proportions for each spot without any reference dataset.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 2/5 · not successfulReplay →
Q36Segment individual cells from the H&E image.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 4/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 4/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 4/5 · successfulReplay →
Q37Perform RNA velocity analysis on this spatial data.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q38Train a deep learning classifier to predict patient survival from this tissue.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 4/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q39Identify spatial niches in this tissue based on cell type composition.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →
Q40Test whether cell types co-localize spatially in this tissue.
STATRun 1 · Claude Sonnet 4.6rated 5/5 · successfulReplay →
STATRun 2 · DeepSeek V4.1 Flash · seed 10rated 5/5 · successfulReplay →
STATRun 3 · DeepSeek V4.1 Flash · seed 20rated 5/5 · successfulReplay →
STATRun 4 · DeepSeek V4.1 Flash · seed 30rated 5/5 · successfulReplay →