The identification of ageing-related proteins is crucial for understanding the molecular mechanisms of human senescence. In this study, we propose a protein classification pipeline based on fine-tuning ESM2, a large-scale protein language model, with Low-Rank Adaptation (LoRA) for the binary classification of ageing-related proteins. Using 320 curated protein sequences from the GenAge database and 320 additional non-ageing proteins from UniProtKB/Swiss-Prot, we constructed a balanced dataset of 640 sequences. After preprocessing and filtering, the training set consisted of 160 positive and 160 negative proteins, while the test set comprised 34 positive and 30 negative proteins. Fine-tuning was performed on ESM2-t33 with a lightweight LoRA configuration targeting attention projections. Evaluation metrics, including accuracy (0.625), F1 score (0.700), and ROC-AUC (0.743), demonstrate the effectiveness of this approach. The results support the feasibility of using transformer-based language models for age-related protein identification and lay the groundwork for further functional clustering and enrichment analyses.