Genome Understanding Evaluation (GUE)
28 datasets/tasks in the native packaging; mapped here to five task families.
- 协议
- 1
- 结果行
- 10
- 工作
- 1
2026-08-06 快照
47 个 benchmark，每一个都保留口径。
查找公开数值结果或可审计缺口。结果行数不等于独立实验数。
47 / 47 个 benchmark
Nucleotide Transformer benchmark
Eighteen downstream tasks are benchmark instances, not eighteen unique task families.
- 协议
- 67
- 结果行
- 821
- 工作
- 1
BEND
Human-genome tasks with biologically aligned splits; inspect each track for leakage controls.
- 协议
- 7
- 结果行
- 105
- 工作
- 1
DART-Eval
Regulatory DNA evaluation across zero-shot, probing and fine-tuning regimes.
- 协议
- 0
- 结果行
- 0
- 工作
- 0
GenBench
Framework and dataset suite; native task count follows current release documentation.
- 协议
- 0
- 结果行
- 0
- 工作
- 0
Genomic Benchmarks
Widely reused classification datasets; simple random-style splits can inflate biological generalization.
- 协议
- 0
- 结果行
- 0
- 工作
- 0
DNALongBench
Long-range genomic tasks; exact sequence length and cell-type packaging vary by track.
- 协议
- 2
- 结果行
- 6
- 工作
- 1
Benchmarking DNA foundation models for genomic and genetic tasks
One-time benchmark study rather than a maintained benchmark suite.
- 协议
- 6
- 结果行
- 8
- 工作
- 1
BEACON
Thirteen native tasks; some reuse OpenVaccine, Optimus and CRISPR datasets.
- 协议
- 13
- 结果行
- 221
- 工作
- 1
RNA-Scope
Board is public; DeepASmRNA and DeeReCT-PolyA/PAS are exposed as distinct tracks; verify exact release and frozen splits before comparative claims.
- 协议
- 79
- 结果行
- 2,093
- 工作
- 1
RNAGym
Three core tasks plus a task-family view of the ribozyme-assay subset; the subset track is separated from the aggregate DMS-effect track.
- 协议
- 2
- 结果行
- 7
- 工作
- 2
Benchmarking pre-trained gLMs for RNA sequence applications
Four native study processes; the GTEx tissue-specific splicing evaluation is additionally mapped to the alternative-splicing task family without changing the native-task count.
- 协议
- 0
- 结果行
- 0
- 工作
- 0
RNA-Puzzles
Blind, rolling CASP-like assessment; puzzle count and target composition change by round.
- 协议
- 50
- 结果行
- 5,657
- 工作
- 1
CASP15 RNA
First dedicated CASP RNA season; preserve target/category filters.
- 协议
- 9
- 结果行
- 312
- 工作
- 1
CASP16 RNA/DNA
Includes RNA and nucleic-acid-containing targets; 42-target figure is reported for the richer RNA set in later assessment work.
- 协议
- 2
- 结果行
- 258
- 工作
- 1
Ribonanza / Stanford Ribonanza RNA Folding
High-throughput DMS/2A3 reactivity prediction; leaderboard is archived.
- 协议
- 18
- 结果行
- 36
- 工作
- 1
OpenVaccine
Same raw competition data is repackaged by several RNA benchmark suites.
- 协议
- 9
- 结果行
- 15
- 工作
- 1
Eterna100
Community-standard inverse-folding set; protocol differences across folding engines reduce comparability.
- 协议
- 1
- 结果行
- 7
- 工作
- 1
NC-Bench
Focused benchmark for non-canonical RNA base pairs.
- 协议
- 0
- 结果行
- 0
- 工作
- 0
CAGI molecular/functional challenges
Only molecular and functional variant challenges are in scope; phenotype-from-exome and patient-level tasks are excluded.
- 协议
- 0
- 结果行
- 0
- 工作
- 0
DREAM5 TF–DNA motif recognition
Historic blinded TF–DNA specificity assessment.
- 协议
- 0
- 结果行
- 0
- 工作
- 0
ENCODE-DREAM in vivo TF binding
Held-out cell types and chromosomes improve realism, but access is through Synapse.
- 协议
- 0
- 结果行
- 0
- 工作
- 0
PRDBv2 / protein–RNA docking benchmark
Docking benchmark; target version, bound/unbound state and redundancy filters must be recorded.
- 协议
- 2
- 结果行
- 4
- 工作
- 1
HARIBOSS / RNA–small-molecule structure benchmark
Useful for binding and docking, but not a single standardized leaderboard.
- 协议
- 0
- 结果行
- 0
- 工作
- 0
R-SIM / RNAmigos2 benchmark
Decoy construction and pocket redundancy are material evaluation choices.
- 协议
- 0
- 结果行
- 0
- 工作
- 0
NucleoBench
Official repository lists 17 design tasks. The ribosomal-loading task is separated as 5′UTR/translation-control design; the remaining instances stay consolidated under cis-regulatory design.
- 协议
- 0
- 结果行
- 0
- 工作
- 0
OligoGym
Twelve curated ASO/siRNA/shRNA datasets with explicit split comparisons.
- 协议
- 24
- 结果行
- 45,932
- 工作
- 1
RNAglib structure–function benchmark
Seven RNA structure–function datasets; task-level splits and structural redundancy controls should be preserved.
- 协议
- 0
- 结果行
- 0
- 工作
- 0
NABench
Large-scale nucleotide fitness benchmark; verify repository release before production adoption.
- 协议
- 0
- 结果行
- 0
- 工作
- 0
Random Promoter DREAM Challenge
Prospective promoter-expression prediction challenge with experimental readout; not counted as a sequence-generation benchmark.
- 协议
- 0
- 结果行
- 0
- 工作
- 0
NucleoDock external benchmark
External set of 125 nucleic-acid–ligand complexes; model-paper benchmark, not community suite.
- 协议
- 0
- 结果行
- 0
- 工作
- 0
SMRTnet RNA–small-molecule benchmark
Emerging benchmark for RNA–small-molecule interaction; check release artifacts and negative construction.
- 协议
- 0
- 结果行
- 0
- 工作
- 0
OmniGenBench
Meta-suite packaging RGB, BEACON, GUE, Genomic Benchmarks and PGB; native_task_count=5 suites, not the sum of their downstream tasks.
- 协议
- 0
- 结果行
- 0
- 工作
- 0
ERNIE-RNA downstream evaluation study
One-time model study spanning 15 named data packages; PrismNet is additionally stratified into 17 HeLa RBP assays.
- 协议
- 12
- 结果行
- 231
- 工作
- 1
Stanford RNA 3D Folding Challenge
Two time-separated competitions with hidden targets; submissions provide five candidate 3D structures and are evaluated best-of-five.
- 协议
- 22
- 结果行
- 82
- 工作
- 1
GenoArmory
Adversarial robustness wrapper over eight GUE histone-mark datasets; four attack families and three defense strategies are evaluation regimes, not new biological task families.
- 协议
- 0
- 结果行
- 0
- 工作
- 0
RNAtive model-quality benchmark
Two released evaluation collections: CASP15 predicted models and Randstr decoys; benchmark package is fully archived on Zenodo.
- 协议
- 0
- 结果行
- 0
- 工作
- 0
EternaBench
Thirty-one chemical-mapping, riboswitch and external thermodynamic datasets; repository is archived but versioned and reproducible.
- 协议
- 8
- 结果行
- 107
- 工作
- 1
AptaBench
6,289 DNA/RNA aptamer–ligand records with fixed stratified, ligand-disjoint and aptamer-disjoint five-fold protocols; anonymous pre-review status should be preserved.
- 协议
- 0
- 结果行
- 0
- 工作
- 0
RNA–RNA Interactions Benchmark
Manually confirmed interaction regions across archaea, bacteria and eukaryotes; the repository moved from UCanCompBio to Gardner-BinfLab.
- 协议
- 7
- 结果行
- 84
- 工作
- 1
RnaBench Riboswitch Design Benchmark
Fixed generative riboswitch design evaluation over 685,109 unique sequences with validity and novelty metrics.
- 协议
- 0
- 结果行
- 0
- 工作
- 0
Deep DNAshape benchmark evaluation
Evaluation against experimental-structure and molecular-dynamics DNA-shape labels; a reusable study protocol rather than a community leaderboard.
- 协议
- 0
- 结果行
- 0
- 工作
- 0
RNP-ΔΔG blind mutation benchmark
Blind and independent RNA–protein affinity tests; only RNA-side mutations are in scope. MS2, diverse-complex and PUM2 tests remain separate protocols.
- 协议
- 3
- 结果行
- 3
- 工作
- 1
ATGC-Gen promoter-generation evaluation
100,000 generated 1,024-bp promoters evaluated with the Sei H3K4me3 activity oracle, MSE, KS and fluency; comparisons are restricted to the same source table.
- 协议
- 1
- 结果行
- 15
- 工作
- 1
RDesign tertiary RNA inverse-design benchmark
Fixed tertiary-structure-to-sequence design evaluation released with code and benchmark data; computational structural recovery only unless separately validated.
- 协议
- 0
- 结果行
- 0
- 工作
- 0
LinearDesign mRNA co-design evaluation
Multi-objective mRNA design with prospective wet-lab measurements of stability, expression and immunogenicity; it must not be ranked with computational-only protocols.
- 协议
- 0
- 结果行
- 0
- 工作
- 0
RhoDesign light-up aptamer evaluation
Structure-guided light-up aptamer generation with prospective fluorescence validation; secondary mapping to tertiary RNA design is recorded separately.
- 协议
- 0
- 结果行
- 0
- 工作
- 0
没有符合当前筛选条件的 benchmark。