Protein Representation Learning

Protein language models are trained primarily with masked language modeling (MLM), which predicts masked amino-acid identities. Joint-embedding predictive architectures (JEPA) instead predict latent representations, but have not been applied to proteins. ProteinJEPA supplements MLM with a cosine loss for predicting the half-depth hidden states of a teacher given the unmasked sequence. On 19 tasks, with ESM2 at 35M and 150M parameters and three pretraining seeds, MLM+JEPA outperforms compute-matched and step-matched MLM-only continued training in 78 and 76 of 114 comparisons (14 losses, 22 ties). The median compute-matched gain is $+0.0106$ on structure- and homology-sensitive tasks versus $+0.0041$ elsewhere, led by SCOPe-40 retrieval and remote homology with improvements of 6.1 percentage points in Recall@1 and 2.7 points in accuracy, respectively. Gains on these tasks increase with model size from 8M to 150M. Against the off-the-shelf checkpoint, MLM+JEPA wins 81 of 114 comparisons (median $+0.0068$) without improving MLM loss. In random initialization the gain is smaller and replicates inconsistently across seeds ($p{=}0.059$). The same recipe improves the causal ProGen3 model, beating a compute-matched next-token-prediction control on 12 of 16 tasks. Ablations show that cosine loss beats mean squared error, while adding shallower targets removes most of the task gain. JEPA-only training collapses downstream performance: latent prediction complements MLM rather than replacing it. Code: https://anonymous.4open.science/r/protJepa-FF24

Protein language models (pLMs) produce per-residue representations that capture evolutionary and structural information, yet their mean-pooled sequence embeddings are not explicitly trained to reflect functional, evolutionary or structural similarity between proteins. We present Protein Sentence Transformers (ProtSent), a contrastive fine-tuning framework for adapting PLMs into general-purpose embedding models. ProtSent trains with MultipleNegativesRankingLoss across five protein-pair datasets: Pfam families, structurally derived hard negatives, AlphaFold DB structural pairs, and StringDB protein--protein interactions, and Deep Mutational Scanning data. We evaluate on 23~downstream tasks using frozen embeddings with a k-nearest-neighbor probe to measure embedding neighborhood quality. On ESM-2 150M, ProtSent improves 15 of 23 tasks, with gains of +105% on remote homology detection, +17% on variant effect prediction, and +19.9% Recall@1 on SCOPe-40 structural retrieval. The 35M variant improves 16 of 23 tasks with +40.5% on remote homology and +15.5% Recall@1 on SCOPe-40. Contrastive fine-tuning restructures the embedding space to better capture protein function and structure, without any task-specific supervision. We release the models, public data, and training recipe and code.

Protein language models are increasingly used to guide experimental and clinical decisions, yet it is often unclear whether a confident prediction reflects recognition of biological evidence or retrieval of a statistical default. We examine this distinction for a near-universal biological rule, that proteins begin with methionine, by tracing the computational pathway through which ESM2-8M produces this prediction. The model does not detect methionine at the masked position. Instead, it retrieves a methionine-favouring signal from a reference representation at the beginning-of-sequence token via a position-specific query assembled across layers, with the final output emerging through competition with context-dependent circuits. To understand how positional information reaches the readout, we introduce a norm-direction decomposition of attention scores within rotary frequency bands. Positional encoding operates through coupled changes in query norm and angular alignment distributed across these bands. On sequences whose true N-terminus is not methionine, where the biological question matters, the model predicts methionine anyway. This is not a correct prediction produced by an unexpected mechanism, but the output of a positional-prior retrieval circuit that matches the statistical average and fails where biology diverges from it. Distinguishing the two requires resolution at the level of individual circuits, frequency bands, and query composition, suggesting that mechanistic verification will be necessary, and challenging, for predictions where the biological stakes are higher. Even for the simplest biological rule, the model's prediction is mediated by a distributed computational circuit rather than direct recognition, suggesting that increasing task complexity will further obscure the relationship between model confidence and underlying biological evidence.

Proteins are inherently multiscale physical systems whose functional properties emerge from coordinated structural organization across multiple spatial resolutions, ranging from atomic interactions to global fold topology. However, existing protein representation learning methods typically operate at a single structural level or treat different sources of structural information as parallel modalities, without explicitly modeling their hierarchical relationships. We introduce PRIME (Protein Representation via Physics-Informed Multiscale Equivariant Hierarchies), a unified framework that models proteins as a nested family of five physically grounded structural graphs spanning surface, atomic, residue, secondary-structure, and protein levels. Adjacent levels are connected through deterministic, physics-informed assignment operators, enabling bidirectional information exchange via bottom-up aggregation and top-down contextual refinement. Experiments on standard protein representation learning benchmarks demonstrate strong and competitive performance across diverse tasks, with particularly notable gains on the Fold Classification benchmark, where PRIME outperforms the strongest geometric GNN baseline by margins of 13.80 and 18.30 points on the harder Superfamily and Fold splits, and achieves a state-of-the-art accuracy of 84.10% on Reaction Class prediction, surpassing all baseline methods, including ESM. Ablation studies confirm that each structural level contributes complementary and non-redundant information, and adaptive cross-attention analysis reveals that PRIME autonomously identifies the most task-relevant structural resolutions at prediction time. Our source code is publicly available at https://github.com/HySonLab/PRIME

Modeling protein sequences as a language has made language models a powerful tool in computational biology, yet the language itself remains poorly understood. A key step toward understanding it is identifying its constituent units. In natural languages, morphemes can occur in multiple forms; similarly, in proteins, mutations can give rise to variations of a unit that persist through evolution, forming families of related units. We introduce PUMA (Protein Units via Mutation-Aware Merging), an algorithm that learns protein units from sequence and explores their mutational variants using substitution matrices, forming a genealogy of unit families. Our results show that mutations remaining within a PUMA family are more often benign than the substitution matrix alone predicts, and that PUMA genealogy improves molecular function representations compared to treating units independently. A case study of a unit family demonstrates relatedness beyond homology. PUMA achieves competitive performance on downstream tasks when used as a protein language model tokenizer. Moreover, collapsing units into families results in a smaller embedding table and faster training. Together, these results support PUMA as a biologically grounded protein vocabulary that organizes protein units into plausible families of mutational variants. The source code is available at https://github.com/boun-tabi-lifelu/PUMA.

\textbf{Motivation:} Pre-trained protein language models (PLMs) such as ESM-2 have become the default representation for biological sequence tasks, but they are computationally heavy and require GPUs both for embedding and for fine-tuning. Whether they are actually necessary for sequence \emph{classification}, as opposed to structure prediction, is rarely tested against strong, principled, lightweight alternatives. This question has direct practical stakes for large-scale genomic surveillance, where embedding millions of sequences on commodity hardware is a recurring bottleneck.\ \textbf{Results:} We introduce Murmur2Vec, an alignment-free, training-free embedding that aggregates $k$-mer counts into a small hash table via the deterministic MurmurHash function, and we cast it as a randomized sketch of the classical $k$-mer spectrum kernel. We provide a complete theoretical treatment: closed-form bias/variance of the inner product, an unbiased signed variant with a Johnson--Lindenstrauss-type concentration bound, an excess-risk bound for downstream linear classifiers that makes the bias--variance trade-off in the hash-table size explicit, and an implicit-regularization mechanism by which collisions damage frequent non-discriminative $k$-mers more than rare lineage-defining ones. Across four classification tasks, SARS-CoV-2 spike lineage (22 classes), HIV-1 Env subtype (8 classes), and two protein-family benchmarks (8 and 6 classes), Murmur2Vec matches a LoRA-fine-tuned 650M-parameter ESM-2 model on the two tasks for which LoRA fine-tuning was run to convergence (SARS-CoV-2 and HIV-1) and ties frozen ESM-2 on the two protein-family tasks, and it \emph{outperforms} the fine-tuned model on the hardest task (SARS-CoV-2 lineage: $0.854$ vs.\ $0.807$ accuracy; macro-F1 $0.684$ vs.\ $0.401$).