Tag: protein-language-models/representations

Apr 21, 2026

Distance between PLM representations of two proteins correlates with functional dissimilarity

protein-language-models/representations

Apr 21, 2026

Distance between averaged PLM embeddings does not correlate with structural difference

Apr 21, 2026

Enzymes can be miniaturized with Monte Carlo sampling and embedding similarity of catalytic residues

protein-language-models/representations

Apr 21, 2026

HMMs cannot identify remote homologs

Apr 21, 2026

High-confidence predictions from protein language models co-cluster together in embedding space and correlate with performance on variant effect prediction tasks

Apr 21, 2026

Larger PLMs are better at homolog detection

Apr 21, 2026

Mean-pooled embeddings outperform other zero-shot approaches for transfer learning of PLMs using the full sequence

protein-language-models/representations

Apr 21, 2026

Membrane proteins are predicted by PLMs via solvent-exposed hydrophobic residues

protein-language-models/representations

Apr 21, 2026

Optimal transport outperforms mean-pooling on property prediction tasks

protein-language-models/representations

Apr 21, 2026

PLM attention maps from specific heads can be used to predict allosteric networks

Apr 21, 2026

PLM attention matrices correspond to 3D contacts

Apr 21, 2026

PLM embeddings contain enough information to align proteins without fine-tuning

Apr 21, 2026

PLM embeddings fine-tuned using contrastive learning outperform other representations in drug-target interaction prediction

Apr 21, 2026

PLM-based sequence searches outperform sequence- and matches structure-based search methods

Apr 21, 2026

PLM-derived antibody representations can distinguish engineered from human-derived Abs

Apr 21, 2026

PLMs learn family-specific protein contacts from sequence context windows of about 20-40 amino acids

protein-language-models/representations

Apr 21, 2026

Protein language model embeddings are more predictive of homology than catalytic efficiency

Apr 21, 2026

Protein language models are able to predict epistasis in a zero-shot setting following a nonlinear transform

Apr 21, 2026

Protein language models are better zero-shot predictors for ranking closely related sequences than distantly related sequences

protein-language-models/representations

Apr 21, 2026

Protein language models can predict zero-shot which proteins belong to the same species

protein-language-models/representations

Apr 21, 2026

Protein language models learn structure-level features, including disorder, in later layers

protein-language-models/representations

Apr 21, 2026

Protein property prediction using PLMs does not benefit from scale except when predicting inferring features of either structural or sparsely populated sequence families

Apr 21, 2026

Sequence homology composition can affect performance of fine-tuned protein language models for variant effect prediction

protein-language-models/representations

Apr 21, 2026

Sequences with lower log-likelihoods are worse for zero-shot variant effect prediction using PLMs

protein-language-models/representations

Apr 21, 2026

Sparse autoencoder-derived features do not outperform PLM-derived embeddings for downstream prediction

protein-language-models/representations

Apr 21, 2026

Structural classifications can be learned from PLM-based alignment prediction

Apr 21, 2026

Structure-based methods outperform sequence-based methods at zero-shot prediction of binding, whereas the reverse is true for zero-shot prediction of enzymatic activity

Apr 21, 2026

Structure-derived embeddings outperform sequence-derived embeddings on LM-based alignment

Apr 21, 2026

Variant effect prediction with homology-aware PLMs improves with ensembling of multiple prompts

protein-language-models/representations

Apr 21, 2026

Zero-shot performance of PLMs, but not inverse folding models, correlates with number of homologs available for training

protein-language-models/representations

Apr 21, 2026

Zero-shot protein fitness prediction using sequence-only NNs can be improved by averaging predictions from many orthologs

protein-language-models/representations

Apr 21, 2026

Concatenating sequence and structural features is not effective for sequence recovery

protein-language-models/representations

Apr 21, 2026

Contrastive learning of PLM embeddings on functional annotation improves variant effect prediction and homolog detection

Apr 21, 2026

Contrastive learning on whole structures leads to learning of distinct substructures

Apr 21, 2026

Correlation between sequence log-likelihood and variant effect prediction performance breaks down as PLMs get larger

protein-language-models/representations

Quartz 4

Explorer

Tag: protein-language-models/representations

Distance between PLM representations of two proteins correlates with functional dissimilarity

Distance between averaged PLM embeddings does not correlate with structural difference

Enzymes can be miniaturized with Monte Carlo sampling and embedding similarity of catalytic residues

HMMs cannot identify remote homologs

High-confidence predictions from protein language models co-cluster together in embedding space and correlate with performance on variant effect prediction tasks

Larger PLMs are better at homolog detection

Mean-pooled embeddings outperform other zero-shot approaches for transfer learning of PLMs using the full sequence

Membrane proteins are predicted by PLMs via solvent-exposed hydrophobic residues

Optimal transport outperforms mean-pooling on property prediction tasks

PLM attention maps from specific heads can be used to predict allosteric networks

PLM attention matrices correspond to 3D contacts

PLM embeddings contain enough information to align proteins without fine-tuning

PLM embeddings fine-tuned using contrastive learning outperform other representations in drug-target interaction prediction

PLM-based sequence searches outperform sequence- and matches structure-based search methods

PLM-derived antibody representations can distinguish engineered from human-derived Abs

PLMs learn family-specific protein contacts from sequence context windows of about 20-40 amino acids

Protein language model embeddings are more predictive of homology than catalytic efficiency

Protein language models are able to predict epistasis in a zero-shot setting following a nonlinear transform

Protein language models are better zero-shot predictors for ranking closely related sequences than distantly related sequences

Protein language models can predict zero-shot which proteins belong to the same species

Protein language models learn structure-level features, including disorder, in later layers

Protein property prediction using PLMs does not benefit from scale except when predicting inferring features of either structural or sparsely populated sequence families

Sequence homology composition can affect performance of fine-tuned protein language models for variant effect prediction

Sequences with lower log-likelihoods are worse for zero-shot variant effect prediction using PLMs

Sparse autoencoder-derived features do not outperform PLM-derived embeddings for downstream prediction

Structural classifications can be learned from PLM-based alignment prediction

Structure-based methods outperform sequence-based methods at zero-shot prediction of binding, whereas the reverse is true for zero-shot prediction of enzymatic activity

Structure-derived embeddings outperform sequence-derived embeddings on LM-based alignment

Variant effect prediction with homology-aware PLMs improves with ensembling of multiple prompts

Zero-shot performance of PLMs, but not inverse folding models, correlates with number of homologs available for training

Zero-shot protein fitness prediction using sequence-only NNs can be improved by averaging predictions from many orthologs

Concatenating sequence and structural features is not effective for sequence recovery

Contrastive learning of PLM embeddings on functional annotation improves variant effect prediction and homolog detection

Contrastive learning on whole structures leads to learning of distinct substructures

Correlation between sequence log-likelihood and variant effect prediction performance breaks down as PLMs get larger