Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-1133
version: 2
kind: revised
reason: Copy rewrite: regenerated from the full source text (Europe PMC PMC13579901) for specificity; removed remarks about truncated source text.
timestamp: 2026-11-04T09:00:01.432388Z
status: published
lens: g_space
sector: science
headline: Genolator enables protein function interpretation using a multimodal large language model fusing genomic and structural interpretation with natural language interaction
dek: Background Decoding the genetic code to unveil its genome functionality is a monumental task which would greatly advance the understanding of disease mechanisms and development of targeted treatments. Although large language models (LLMs) have transformed natural language processing across diverse domains, translating the complex language of DNA into human-readable form remains challenging due to genomic data complexity and unexplored regions of the human genome. Current language models either are capable of pro…
gain_title: On a holdout test set of 34,743 questions, the fully fine-tuned Genolator reached an F1-score of about 0.98 and an MCC of 0.959 on confirm-or-deny protein function questions, versus 0.913 for an XGBoost classifier and 0.179 for GPT 4.1.
problem_title: (none)
trace_subject: (none)
gain_reading: On a holdout test set of 34,743 questions, the fully fine-tuned Genolator reached an F1-score of about 0.98 and an MCC of 0.959 on confirm-or-deny protein function questions, versus 0.913 for an XGBoost classifier and 0.179 for GPT 4.1.
gain_evidence: Genolator (full; Evo2 & PST), in which both the virtual token projectors and the Llama model were jointly trained/fine-tuned, achieved an overall F1-score of around 0.98. | GPT 4.1 [ 38 ] reached a MCC of 0.179 and the XGBoost model reached a MCC of 0.913.
problem_reading: (none)
problem_evidence: (none)
quick_read: Researchers built Genolator, a multimodal model based on Llama that combines DNA embeddings from Evo2, amino acid embeddings from ESM2 or structure-aware embeddings from PST, and an English question. It was fine-tuned on more than 365,000 question-answer pairs built from Gene Ontology terms covering subcellular localization, molecular function and biological processes. On 34,743 held-out questions about unseen proteins, the best version scored an F1 of about 0.98 and an MCC of 0.959.

General-purpose models did far worse. GPT 4.1 reached an MCC of only 0.179. In a case study it misidentified BRCA1 as TP53 and HTT and IHH as COL1A1 and MPZ, and it claimed to have run BLAST searches whose results were fabricated or showed no significant similarity. A tuned XGBoost classifier on PST and BioBERT embeddings reached an MCC of 0.913, so Genolator's edge over the specialist baseline was modest.

Open-ended answers were less strong. Predicted terms overlapped the ground truth for 78.51% of biological process questions, 89.19% for cellular component and 89.59% for molecular function, but exact matches were only 2.51% to 5.83%. The authors present the tool as a way to make genomic embeddings accessible through natural language for research on uncharacterized proteins, not as a validated clinical tool.
limitation: Open-ended answers were judged by a GPT 4.1-based automatic mapping rather than experts, exact matches were rare, and the model was not hyperparameter-tuned or tested across multiple species.
tag: Evidence-backed gain
key_points: Genolator was fine-tuned on over 365,000 question-answer pairs, including both genuine associations and deliberately constructed denial questions. | The holdout test set had 34,743 questions, and a KMeans-based split placed test proteins in distinct regions of the multimodal embedding space to mimic proteins of unknown function. | The Evo2 + ESM2 variant reached an F1 of 0.97 and an MCC of 0.948; the projector-only ablations reached MCCs of 0.895 (ESM2) and 0.924 (PST). | GPT 4.1 identified the INS (insulin) sequence correctly but misidentified the longer BRCA1 and IHH sequences and described BRCA1 as involved in transport across membranes. | Untrained Llama-3 8B could not be benchmarked because long DNA inputs exceeded memory and time limits even on an NVIDIA H100, and its answers included amino-acid gibberish and irrelevant follow-up questions. | Exact matches on open-ended questions were only 5.44% for biological process, 5.83% for cellular component and 2.51% for molecular function. | Biological process was the hardest aspect; for example, regulation of DNA-templated transcription was generated 97 times against 444 in the ground truth. | Zeroing the PST tokens raised test loss by 3.31% in the Evo2 + PST model, versus 0.56% when Evo2 tokens were removed.
rundown: Genolator is a finetuned Llama model that receives an English question plus embeddings of a gene's DNA sequence (Evo2), its amino acid sequence (ESM2) and, in the best variant, a structure-infused protein embedding (PST). Token projectors fuse these inputs, and the model replies in English. Answers were derived from three Gene Ontology aspects, which a GPT 4.1 model rephrased into plain language for training.

In the confirm-or-deny task the best model reached an MCC of 0.959 and an F1 of about 0.98, ahead of the XGBoost classifier at 0.913 and GPT 4.1 at 0.179. Across categories, molecular function was easiest and biological process hardest. The projector-only version outperformed GPT 4.1 but was only slightly different from XGBoost.

In the GPT 4.1 case study, the insulin gene was handled correctly, but for BRCA1 and IHH the model reported identifications with no significant sequence similarity to the reference. The authors note that GPT 4.1 likely saw many genomic sequences in training yet still could not interpret them in context.

Hidden-state analysis showed confirmed and denied associations separating clearly, and related terms clustering together, such as nuclear localizations and enzyme activities. Attention analysis showed Evo2 tokens receiving mean attention above 0.6 in most layers from 2 to 31 in the ESM2 model, and the token ablation showed the structure-aware PST tokens mattered more than Evo2 tokens in the PST model.

The authors say open-ended evaluation relied on GPT 4.1 as a judge to map answers to GO Slim terms because expert review of thousands of predictions was not feasible. Hyperparameters of the token projectors and Llama model were not tuned, and further work is needed for multiple species.
sources:
- peer_reviewed | Genome Biology | https://doi.org/10.1186/s13059-026-04274-w | 2026-09-16
prev: b6af1a977d29f4fd939064c5e325627637fdb51ff010d4ac1abb3b75ea16230c

sha256

3be5a823134d791121304f579a10d6f812458d36559c03ddbddcd790eccd15e4

previous

b6af1a977d29f4fd939064c5e325627637fdb51ff010d4ac1abb3b75ea16230c