weijiang99/clinvar-evidence-conclusion-classifier

ClinVar Evidence/Conclusion/Description Classifier

A fine-tuned bert-base-uncased model for classifying sentences in ClinVar submission comments into three categories: description, evidence, or conclusion.

Model Description

This model is used to preprocess ClinVar submission comments by identifying and filtering out sentences that contain conclusions and descriptions, retaining only sentences that convey evidence for downstream analysis.

Base model: bert-base-uncased Task: Single-label sequence classification (3 classes)

Labels

ID	Label	Meaning
0	LABEL_0	Evidence
1	LABEL_1	Description
2	LABEL_2	Conclusion

Usage

from transformers import BertForSequenceClassification, BertTokenizer
import torch

model = BertForSequenceClassification.from_pretrained("weijiang99/clinvar-evidence-conclusion-classifier")
tokenizer = BertTokenizer.from_pretrained("weijiang99/clinvar-evidence-conclusion-classifier")

model.eval()
sentences = ["The variant was observed in 3 affected family members.", "This variant is likely pathogenic."]
inputs = tokenizer(sentences, padding=True, truncation=True, return_tensors="pt")

with torch.no_grad():
    outputs = model(**inputs)
    predictions = torch.argmax(outputs.logits, dim=1)

label_map = {0: "evidence", 1: "description", 2: "conclusion"}
for sentence, pred in zip(sentences, predictions):
    print(f"{label_map[pred.item()]}: {sentence}")

Intended Use

Preprocessing ClinVar variant interpretation comments to extract evidence sentences for downstream natural language processing tasks.