Aug 2026· International journal of life sciences and biotechnology· Vol 9, pp. 90-105· 0 citations· 38 references
TL;DR
Attention weight analysis shows the model identifies biologically relevant signals, including C-terminal membrane regions and N-terminal mitochondrial sequences, confirming that ESM-2 embeddings enable effective performance with simplified architectures.
Abstract
Predicting protein subcellular locations computationally is crucial for analyzing large protein datasets. A key issue is that similar sequences in training and test sets artificially inflate accuracy estimates. This study investigates whether Protein Language Model (PLM) features alone can achieve strong predictions using simple classifiers instead of complex architectures. We developed a streamlined deep learning framework combining pre-trained ESM-2 embeddings with an attention-enhanced Bi-LSTM network, deployed as a 3-fold ensemble with soft voting. Training used eukaryotic sequences with ≤40% similarity to ensure rigorous evaluation. The model achieved 86.81% accuracy (MCC = 0.825) on test data—a +22.27% improvement over an SVM baseline (64.54%, MCC = 0.530). On 86 newly released 2024 proteins, the system reached 88.37% accuracy (MCC = 0.827), surpassing DeepLoc 2.1 (80.23%, MCC = 0.714, p=0.007) and MULocDeep (77.91%, MCC = 0.682, p=0.019). However, the small validation set (N=86) and limited representation in categories like Mitochondrion (N=5) require cautious interpretation. The method only handles single-location assignments across four compartments, excluding multi-location proteins. Attention weight analysis shows the model identifies biologically relevant signals, including C-terminal membrane regions and N-terminal mitochondrial sequences, confirming that ESM-2 embeddings enable effective performance with simplified architectures.
Binary protein classification supports diverse tasks in computational biology, including pathway-membership inference and sequence-based candidate prioritization. Protein language models generate information-rich residue-level representations, but downstream classifiers commonly compress them using fixed pooling operations that may discard task-relevant sequence context. We present ETAP-CLF, a compact framework that combines pretrained per-residue ESM3 embeddings with lightweight transformer contextualization and learned attention pooling to classify variable-length proteins and generate residue-level attention scores. The ESM3 parameters remained frozen, and the same ETAP-CLF architecture and hyperparameter configuration were used across ferroptosis-, senescence-, and pyroptosis-associated protein prediction. ETAP-CLF achieved AUROCs of 0.98, 0.95 and 0.91 for these tasks, respectively. In the ferroptosis benchmark, ETAP-CLF outperformed the evaluated published models. These results demonstrate that a common downstream design can adapt to multiple process-associated classification tasks without fine-tuning the task-specific model architecture. ETAP-CLF provides a generalizable approach for sequence-based protein prioritization and a basis for broader evaluation across binary protein-classification problems.
Jianyu Ren, Hanli Jiang, Puchangxin Li et al.· bioRxiv· 0 citations
Over 99.9% of known protein sequences lack experimentally validated functional annotations. We present ProtJEPA, a multimodal Joint-Embedding Predictive Architecture that trains a sequence-only student encoder to predict joint embeddings spanning ten biological modalities—sequence, structure, knowledge graph, protein interactions, literature, localization, tissue expression, GO function, anatomy, and disorder—requiring only sequence at inference. The key innovation is target whitening, which eliminates severe anisotropy in joint targets (mean cosine 0.984 to 0.086) and prevents representation collapse without covariance regularization. On 1,828 held-out dark proteins with zero primary Pfam family overlap with training, ProtJEPA achieves 58.07% Hit@10 on zero-shot GO retrieval (+2.80 pp, p = 0.020), 69.99% enzyme class accuracy (+9.64 pp, p < 0.001), and +11.87 pp subcellular localization at 1% labels (p < 0.001). Under realistic dark-protein deployment conditions where relational modalities are unavailable, ProtJEPA significantly outperforms naive concatenation of remaining modalities. Cross-domain evaluations on drug–target interaction and disorder prediction confirm transfer beyond training modalities, with the T1-only < ESMC < ProtJEPA ordering replicated across six independent tasks. Ablations establish that Phase 1 aggregator pretraining and target whitening are each independently load-bearing.
Vaibhava Lakshmi Ravideshik, Jinha Kim, M. Kellis· bioRxiv· 0 citations
Results support the use of pretrained protein language model embeddings as an effective foundation for enzyme annotation by combining large-scale sequence representations with a lightweight supervised classifier and may facilitate functional annotation of protein sequences derived from large genomic and metagenomic datasets.
This study evaluates four encoding schemes—one-hot, k-mer (substring-based encoding), embeddings, and Position-Specific Scoring Matrix (PSSM) using Convolutional Neural Networks (CNNs) and Long Short-Term Memories (LSTMs) and shows that k-mer encoding achieved the highest accuracy.
T. Kurniawan, Deshinta Arova Dewi, Randy Joy Magno Ventayen· Journal of Advances in Infor...· 0 citations
Overall, DrugPLMFormer provides a reproducible, leakage-aware framework for retrospective sequence-based druggability screening and target prioritization, while prospective validation and experimental confirmation remain necessary before operational deployment.
Z. Kafi, Khosro Rezaee, Hossein Eslami· Journal of King Saud Univers...· 0 citations
Findings show that performance estimates in protein function prediction should be interpreted as outcomes of complete data-centric workflows rather than isolated properties of predictive models.
