TrendBio

#4 · Proteomics

Benchmarking computational omics methods against defined references and scenarios

11 preprints · 3 published · ↑ 1.6× vs 4-wk avg
Items per day, last 8 weekstoday: 14

The dominant thread is systematic evaluation of computational methods under controlled conditions. scArchon provides a modular Snakemake framework benchmarking nine single-cell perturbation response prediction tools across six scRNA-seq datasets. BatchVaria offers a variance-aware evaluation of batch correction in SummarizedExperiment objects, tested on TCGA breast cancer data. A siRNA efficacy benchmark removes cross-validation leakage and compares z-curve RNA encodings, reporting AUC 0.845 on leakage-free validation. A unified benchmark of synthetic data generators for clinical transcriptomic cancer cohorts weighs biological fidelity, statistical similarity and attack-based evaluation. simCRISPR models experimental complexity in pooled screens, benchmarking GxE analysis, normalization and log2FC thresholds. A causal transportability framework with a semi-synthetic Perturb-seq simulator shows that deep models often fail to generalize across settings. What is new is anchoring comparisons to explicit scenario design and leakage-free or external validation rather than single-dataset performance.

Published papers

Preprints

Topic history

DateRank
#4
#4
#3
#1
#1
#1
#1
#4
#1
#4
#1
#1
#2
#3
#3
#3
#2
#1
#4
#4
#1
#3
#2
#4
#5