Experiments

Searchable full-text extractions: founding hypothesis, core claims, experimental setups, key results and statistics — pulled out of each paper as structure. Search a cell line, an assay or an entity (e.g. HUH7) and find every paper that worked with it. This corpus stands on its own: most entries carry no reproduction assessment (yet).

• Advancing codon language modeling with synonymous codon constrained masking.

  PMID 41736545 · PMC12956333 · Nucleic acids research · 2026 · 8 claims · 7 setups

  Full-text index only

  SynCodonLM introduces synonymous codon-constrained masking, restricting masked-codon prediction to only synonymous codon options via logit masking before softmax

  cds syncodonlm escherichia rscu masked t-sne pca

• Mouse-Geneformer: A deep learning model for mouse single-cell transcriptome and its cross-species utility.

  PMID 40106407 · PMC11964219 · PLoS genetics · 2025 · 7 claims · 6 setups

  Has reproduction · 95

  Mouse-Geneformer, a Transformer Encoder model pre-trained via masked-token self-supervised learning on mouse-Genecorpus-20M, was successfully constructed following the original human Geneformer architecture.

  healthy scrna-seq disease scrna-seq type perturbation

• pmid-41555433

  PMID 41555433 · PMC12903630 · 8 claims · 8 setups

  Full-text index only

  CLAMP, a chromatin-accessibility language model, predicts protein-mediated chromatin loops across 10 species, 18 proteins, and 24 cell types with superior performance versus existing methods.

  species gm12878 types chromatin cross-protein cross-cell-type

• Separating selection from mutation in antibody language models.

  PMID 41944291 · PMC13056363 · eLife · 2026 · 8 claims · 6 setups

  Full-text index only

  Masked antibody language models such as AbLang2 are biased by nucleotide-level mutation processes (germline memorization, codon table, SHM rate variation)

  single nine antibody reconstructed masked-site correlation functional

• Using ESTs to improve the accuracy of de novo gene prediction.

  PMID 16817966 · PMC1534067 · BMC bioinformatics · 2006 · 8 claims · 8 setups

  Full-text index only

  TWINSCAN_EST combines EST alignments with TWINSCAN via a trainable 'ESTseq' representation and improves exact gene structure prediction accuracy on the whole C. elegans genome

  elegans genome chromosome novo gene partial

• Retentive Network promotes efficient RNA language modeling of long sequences.

  PMID 41814064 · PMC13111708 · Communications biology · 2026 · 8 claims · 6 setups

  Full-text index only

  RNAret, a RetNet-based RNA language model with O(n) complexity, achieves training parallelism and low computational overhead while processing long RNA sequences

  million rnacentral mirtarraw rnastralign self-supervised t-sne mirna-mrna

• Comprehensive benchmarking of large language models for RNA secondary structure prediction.

  PMID 40205851 · PMC11982019 · Briefings in bioinformatics · 2025 · 7 claims · 4 setups

  Has reproduction · 88

  Existing RNA-LLMs had not previously been evaluated for secondary structure prediction in a unified, fair experimental setup with the same datasets and prediction model.

  archiveii bprna large-scale secondary cross-family self-supervised

• Pegasys: software for executing and integrating analyses of biological sequences.

  PMID 15096276 · PMC406494 · BMC bioinformatics · 2004 · 8 claims · 7 setups

  Full-text index only

  Pegasys is a flexible, modular, customizable software system for executing and integrating heterogeneous biological sequence analysis tools

  eukaryotic formatted genomic input initio sequence gene

• BiRNA-BERT allows efficient RNA language modeling with adaptive tokenization.

  PMID 41266599 · PMC12635123 · Communications biology · 2025 · 8 claims · 8 setups

  Has reproduction · 50

  BiRNA-BERT uses adaptive dual-tokenization that dynamically selects nucleotide-level (NUC) or byte-pair encoding (BPE) tokens based on input sequence length

  rfam plant lncrna-mirna sequences unsupervised mirna-lncrna length-binned

• Predicting failure rate of PCR in large genomes.

  PMID 18492719 · PMC2441781 · Nucleic acids research · 2008 · 7 claims · 8 setups

  Full-text index only

  The number of predicted primer-binding sites in genomic DNA is the most important factor determining PCR failure.

  genomic genome pcr combined large-scale exact-match mismatch-tolerant

• The Vertebrate Genome Annotation (Vega) database.

  PMID 15608237 · PMC540089 · Nucleic acids research · 2005 · 8 claims · 8 setups

  Full-text index only

  Vega is a community database for browsing manual annotation of finished vertebrate genome sequences, based on an extended Ensembl-style schema.

  chromosomes genomic manual sequence initio

• First genome assemblies of Neotropical Thoracobombus bumblebees Bombus pauloensis and Bombus pullatus.

  PMID 41436027 · PMC12958814 · G3 (Bethesda, Md.) · 2026 · 7 claims · 8 setups

  Full-text index only

  This study produced the first genome assemblies of Neotropical Bombus (Thoracobombus) species, B. pauloensis and B. pullatus

  pauloensis masked long-read k-mer novo

• Chromosome-scale genome sequencing, assembly and annotation of six genomes from subfamily Leishmaniinae.

  PMID 34489462 · PMC8421402 · Scientific data · 2021 · 8 claims · 8 setups

  Has reproduction · 92

  Chromosome-scale genomes of six Leishmaniinae species (five L. (Mundinia) species and one Porcisia species) were sequenced, assembled and annotated, providing genome, proteome, transcriptome and GFF outputs for taxa previously lacking public reference genomes

  six same polished each nanopore illumina novo