Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
The authors reframe protein functional annotation from rigid, closed-vocabulary multi-label classification into an open-ended sequence-to-text generation task using a 3B-parameter Ministral-3 model. Moving to natural language generation allows models to bypass the expressive constraints of predefined ontologies like Gene Ontology to capture nuanced biochemical context directly from primary sequences. Evaluated via a senior curator-prompted LLM-as-a-judge protocol, the QLoRA-adapted 3B model produces biologically meaningful, curator-style functional descriptions across a substantial subset of proteins despite its compact footprint.
Protein annotation doesn't have to be trapped in rigid ontologies: a 4-bit 3B LLM can generate curator-grade functional descriptions directly from raw amino acid sequences.
Functional annotation of newly sequenced proteins remains a bottleneck in molecular biology: the number of sequences in public repositories grows far faster than the capacity for manual curation. Most computational approaches consider annotation as multi-label classification over a fixed ontology, which constrains predictions to a predefined label set. In this work we study the the protein annotation as a sequence-to-text generation problem. We fine-tune the 3B-parameter Ministral 3 base model with QLoRA (4-bit NF4 quantization with low-rank adapters) on sequence annotation pairs. We assess predictions with an LLM-as-expert protocol: a GPT model prompted as a senior molecular-biology curator scores organism identification as binary and function annotation quality. We conclude that QLoRA-fine-tuned compact LLMs can generate curator-style annotations with genuine biological value for a substantial subset of proteins. We also discuss future directions in data quality, model scaling, and evidence grounding that are needed to make the approach sufficiently reliable for practical use.