Ontology accelerates few-shot learning capability of large language model: A study in extraction of drug efficacy in a rare pediatric epilepsy
Bibliographic record
Abstract
OBJECTIVE: Dravet Syndrome (DS) is a developmental and epileptic encephalopathy that is characterized by severe, prolonged motor seizures and high resistance to multiple antiseizure medications (ASMs) with multiple comorbidities. Evaluating the efficacy of new drugs in DS preclinical models and mapping them to human phenotypes of DS through analysis of published literature is an important goal for improving outcomes in this rare pediatric epilepsy. MATERIALS AND METHODS: Large language models (LLM) have demonstrated great promise in parsing published literature; however, the performance of LLMs falls short in medical applications. In this study, we investigate the effectiveness of domain ontology developed by human experts to optimize LLMs for medical text processing in a rare disease. Utilizing a benchmark dataset that describes the efficacy of 17 ASMs tested in preclinical models and DS patients, we define a new ontology-augmented phased in-context learning (PCL) approach to process 4935 full-text DS articles. We expand this analysis to 7 new drugs that demonstrate efficacy in reducing seizures to identify gaps in current knowledge for designing new experimental studies for drug discovery in DS. RESULTS: Few-shot or in-context learning is a foundational capability of LLMs and the few-shot learning capability of the Gemini 1.0 Pro version LLM dramatically increases when we augment prompts with the DS epilepsy ontology. The DS epilepsy ontology is the largest epilepsy and seizure ontology in clinical use that was developed by DS basic scientists and clinical neurologists. The ontology-augmented PCL prompt achieves 100% accuracy in reproducing the benchmark drug efficacy dataset for 17 ASMs with only two examples for in-context learning. CONCLUSION: The new ontology-augmented PCL approach significantly accelerates the few-shot learning capabilities of the Gemini LLM, thereby reducing the number of required examples and time needed to optimize LLMs for medical applications.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".