benchmarking

The paper evaluates AlphaFold's predictions of transmembrane protein complexes, finding that while most show plausible membrane geometry, some have inconsistencies and limitations in predicting lipid interactions.

The paper develops TriCasRank, a model for ranking CRISPR-Cas12a crRNA activity using sequence features, reference mapping, and historical data, showing improved correlation but comparable performance to an XGBoost model.

The paper introduces WhatRes, a tool to evaluate AI-predicted protein structures by comparing their geometry to experimental structures across resolution and realism.

SPARK quantifies phage-antibiotic interactions using growth curves and shows synergy, antagonism, and dynamic changes across bacterial isolates and time.

The paper introduces LADD to measure lineage path deviations from inferred developmental dynamics in single-cell RNA-seq data and finds lineages move faster than optimal paths.

The paper introduces PRISM to select low-fidelity simulators for multi-fidelity inference in vertex models, finding that early stopping is unreliable while coarser timesteps are effective.

This paper introduces SHARP, a method using cosmological web analysis to characterize brain ageing and cognitive impairment by analyzing MRI data from CamCAN and OASIS-3.

This study evaluates AlphaGenome's ability to predict gene expression changes after promoter perturbations, finding discrepancies between CAGE and RNA-sequencing predictions.

This paper introduces GeneEmbedLLM, a framework using large language models to generate gene embeddings from multi-source textual annotations, achieving competitive performance in various downstream tasks.

BioMeme predicts cell responses to new perturbations by recalling measured experiments, outperforming neural models in accuracy and speed.

This paper evaluates tabular foundation models as species distribution models, finding they perform competitively without per-dataset tuning but are sensitive to biased data.

This paper introduces SSAF-T, a segmentation-free framework for cell-type inference in spatial transcriptomics that avoids cell segmentation and enables accurate cell-type-guided segmentation.

The paper evaluates Dorado and Barbell preprocessing tools for metagenomic reads, finding Barbell reduces barcode-like sequences more effectively while retaining most reads.

This paper introduces hybrid RY-mers to improve aDNA alignment sensitivity and efficiency, showing higher accuracy and reduced runtime compared to existing methods.

This paper introduces MirrorDNA, a DNA language model with a strand-equivariant tokenizer that improves prediction consistency across DNA strands.

The paper introduces a hub labeling method to reduce the space required for distance indexing in large pangenome graphs, improving query support and reducing storage needs.

This paper introduces Braincodec, a deep encoder that improves sound transmission via cortical stimulation, achieving high-fidelity sound reconstruction and matching non-implanted mice in sound discrimination.

This paper introduces pLM-bVAE, a binary variational autoencoder enhanced with protein language model representations, showing improved sequence reconstruction and design performance in quantum annealing-based protein sequence design.

scCircaSim generates circadian single-cell and spatial transcriptomic data with known ground truth to evaluate analysis methods.

The paper shows that plate design confounds donor variation in single-cell perturbation screens and demonstrates how improved analysis methods can change interpretations of recent studies.

This paper introduces a method to increase multiplexing in TaqMan qPCR by encoding target-specific information into amplification curves and decoding them with a transformer model, achieving 8-plex detection with high accuracy.

OncoSpot benchmarks cancer spatial transcriptomics methods across tumor evolutionary modes by simulating realistic cancer data with known subclonal and copy-number features.

SwiftLD, a faster alternative to vcftools for computing LD, shows European LD proxies poorly tag 43% of GWAS leads in African populations.

TIGRdisco identifies TIGR arrays and Tas proteins from nucleotide sequences, recovering most array units and uncovering unannotated loci across datasets.

This study evaluates AlphaFold 3's accuracy and reliability in predicting RNA structures and protein-RNA interfaces, finding that confidence scores do not always correlate with correct predictions.

This paper evaluates the effectiveness of a blinded AI peer review system (BARS) in auditing biomedical literature, finding it improves consistency and detects flaws while highlighting issues with megajournals and template studies.

This paper introduces an end-to-end in silico framework to validate and test clinical genome sequencing pipelines from simulated reads to reportable findings using a fixed truth set.

SPaRTA enables non-experts to query spatial-omics data via natural language, achieving 63-77% accuracy with fine-tuning on cancer tissue datasets.

This paper introduces AutoSPECTRA, a protocol for auditing generalizability of scientific foundation models across biological domains.