arXiv subjects

Nicole Bussola

Publications and source records attributed to Nicole Bussola.

4 recordsLinked to original sources

arXiv2026

JEPA-DNA: Grounding Genomic Foundation Models through Joint-Embedding Predictive Architectures

Genomic Foundation Models (GFMs) typically rely on Masked Language Modeling (MLM) or Next-Token Prediction (NTP) to learn the "Laws of Nature". While effective at capturing local syntax, these generative paradigms prioritize token-level reconstruction over high-level functional context. We introduce JEPA-DNA, a model-agnostic continual training framework that integrates a Joint-Embedding Predictive Architecture (JEPA) with traditional generative objectives. By supervising global sequence embeddings in a latent space, JEPA-DNA forces models to predict the functional representations of masked genomic segments, shifting the learning signal from token recovery to semantic alignment. We evaluate JEPA-DNA on 17 diverse genomic benchmark tasks, demonstrating consistent gains in linear probing and zero-shot performance regardless of the underlying GFM architecture or generative objective. Our framework establishes a new state-of-the-art for GFMs, surpassing the best existing models by bridging generative precision with latent semantic grounding. Through extensive ablation studies, we further characterize the synergistic interplay between generative and latent objectives. Our code is publicly available at https://github.com/NVIDIA-Digital-Bio/JEPA-DNA.

cs.AI↗

arXiv2026

Fine-Tune, Don't Prompt, Your Language Model to Identify Biased Language in Clinical Notes

Clinical documentation can contain emotionally charged language with stigmatizing or privileging valences. We present a framework for detecting and classifying such language as stigmatizing, privileging, or neutral. We constructed a curated lexicon of biased terms scored for emotional valence. We then used lexicon-based matching to extract text chunks from OB-GYN delivery notes (Mount Sinai Hospital, NY) and MIMIC-IV discharge summaries across multiple specialties. Three clinicians annotated all chunks, enabling characterization of valence patterns across specialties and healthcare systems. We benchmarked multiple classification strategies (zero-shot prompting, in-context learning, and supervised fine-tuning) across encoder-only models (GatorTron) and generative large language models (Llama). Fine-tuning with lexically primed inputs consistently outperformed prompting approaches. GatorTron achieved an F1 score of 0.96 on the OB-GYN test set, outperforming larger generative models while requiring minimal prompt engineering and fewer computational resources. External validation on MIMIC-IV revealed limited cross-domain generalizability (F1 < 0.70, 44% drop). Training on the broader MIMIC-IV dataset improved generalizability when testing on OB-GYN (F1 = 0.71, 11% drop), but at the cost of reduced precision. Our findings demonstrate that fine-tuning outperforms prompting for emotional valence classification and that models must be adapted to specific medical specialties to achieve clinically appropriate performance. The same terms can carry different emotional valences across specialties: words with clinical meaning in one context may be stigmatizing in another. For bias detection, where misclassification risks undermining clinician trust or perpetuating patient harm, specialty-specific fine-tuning is essential to capture these semantic shifts. * Equal contribution.

cs.CL↗

arXiv2024

Natural Language Programming in Medicine: Administering Evidence Based Clinical Workflows with Autonomous Agents Powered by Generative Large Language Models

Generative Large Language Models (LLMs) hold significant promise in healthcare, demonstrating capabilities such as passing medical licensing exams and providing clinical knowledge. However, their current use as information retrieval tools is limited by challenges like data staleness, resource demands, and occasional generation of incorrect information. This study assessed the potential of LLMs to function as autonomous agents in a simulated tertiary care medical center, using real-world clinical cases across multiple specialties. Both proprietary and open-source LLMs were evaluated, with Retrieval Augmented Generation (RAG) enhancing contextual relevance. Proprietary models, particularly GPT-4, generally outperformed open-source models, showing improved guideline adherence and more accurate responses with RAG. The manual evaluation by expert clinicians was crucial in validating models' outputs, underscoring the importance of human oversight in LLM operation. Further, the study emphasizes Natural Language Programming (NLP) as the appropriate paradigm for modifying model behavior, allowing for precise adjustments through tailored prompts and real-world interactions. This approach highlights the potential of LLMs to significantly enhance and supplement clinical decision-making, while also emphasizing the value of continuous expert involvement and the flexibility of NLP to ensure their reliability and effectiveness in healthcare settings.

cs.AI↗

arXiv2020

AI slipping on tiles: data leakage in digital pathology

Reproducibility of AI models on biomedical data still stays as a major concern for their acceptance into the clinical practice. Initiatives for reproducibility in the development of predictive biomarkers as the MAQC Consortium already underlined the importance of appropriate Data Analysis Plans (DAPs) to control for different types of bias, including data leakage from the training to the test set. In the context of digital pathology, the leakage typically lurks in weakly designed experiments not accounting for the subjects in their data partitioning schemes. This issue is then exacerbated when fractions or subregions of slides (i.e. "tiles") are considered. Despite this aspect is largely recognized by the community, we argue that it is often overlooked. In this study, we assess the impact of data leakage on the performance of machine learning models trained and validated on multiple histology data collection. We prove that, even with a properly designed DAP (10x5 repeated cross-validation), predictive scores can be inflated up to 41% when tiles from the same subject are used both in training and validation sets by deep learning models. We replicate the experiments for $4$ classification tasks on 3 histopathological datasets, for a total of 374 subjects, 556 slides and more than 27,000 tiles. Also, we discuss the effects of data leakage on transfer learning strategies with models pre-trained on general-purpose datasets or off-task digital pathology collections. Finally, we propose a solution that automates the creation of leakage-free deep learning pipelines for digital pathology based on histolab, a novel Python package for histology data preprocessing. We validate the solution on two public datasets (TCGA and GTEx).

q-bio.QM↗