Jian Sun, Kingshuk Ghosh, Lilianna Houston, Mohammad H. Mahoor: MT-ProtBERT: Multi-task Learning ProtBERT for Intrinsically Disordered Proteins Classification with Scarce Data https://arxiv.org/abs/2609.25334 https://arxiv.org/pdf/2609.25334 https://arxiv.org/html/2609.25334

170M params under MIT is a genuinely useful size for protein work on a single GPU — but the card ships no benchmarks against ESM or ProtBERT, so the capability claim is unverified. Hard to justify switching from ESM until someone publishes a head-to-head.

DeepVIC separates virulence factor identification and classification into distinct modules, using ProtBert-BFD embeddings for binary classification and integrating these with PSSM-derived evolutionary features for multiclass prediction across 14 VFDB categories.

Nah ESM-2, CodonBERT, ProtBERT etc. are all LLMs

🧠 Can post-processing and deep learning push protein prediction to near perfection?

🔗 Post-processing enhances protein secondary structure prediction with second order deep learning and embeddings. DOI: doi.org/10.1016/j.cs...

📚 CSBJ: www.csbj.org

#PSSP #ProteinStructure #ProtBERT #AlphaFold #AI

I want to start with a thank you to sine amazing resources: the Prokaryotic Virus Remote Homologous Groups database (aka. PHROGS) phrogs.lmge.uca.fr doi.org/10.1093/narg... and ProtBert-BFD huggingface.co/Rostlab/prot... ieeexplore.ieee.org/document/947... by the Burkhard Rost lab. (2/9)

RRIDs were included in this in iScience paper. RRIDs improve reproducibility in scientific research. #RRID #OpenScience #STMpublishing

ProtAttn-QuadNet: An attention-based deep learning framework for protein-protein interaction prediction using ProtBERT embeddings https://www.biorxiv.org/content/10.1101/2025.11.17.688786v1

“Insights into the inner workings of transformer models for protein function prediction” has been published

Provides evidence that LLMs (here, ProtBERT) has learned to distinguish transmembrane regions from other parts of a protein

doi.org/10.1093/bioi...
github.com/markuswenzel...

“GPCR-BERT: Interpreting Sequential Design of G Protein Coupled Receptors Using Protein Language Models” 🧪🧶🧬

Looking at how exactly ProtBERT represents GPCRs and their sequence motifs when fine-tuned on their sequences

arxiv.org/abs/2310.19915
github.com/Andrewkimmm/...

“PGraphDTA: Improving Drug Target Interaction Prediction using Protein Language Models and Contact Maps” 🧶🧬

Some interesting comparisons between ProtBERT and DistilProtBERT on DTI tasks here

arxiv.org/abs/2310.04017
anonymous.4open.science/r/PGraphDTA

"cdsBERT - Extending Protein Language Models with Codon Awareness"

This codon-based PLM outperforms ProtBERT on EC number prediction

www.biorxiv.org/content/10.1...

ProtAttn-QuadNet: An attention-based deep learning framework for protein-protein interaction prediction using ProtBERT embeddings https://www.biorxiv.org/content/10.1101/2025.11.17.688786v1

BERT-T6: Towards High-accuracy T6SS Bacterial Toxin Identification Using Protein Language Model [new]
T6SE ID via transfer learning & ProtBert embeds.