Biology subjects

Meehl, J.

Publications and source records attributed to Meehl, J..

3 recordsLinked to original sources

bioRxiv2026

EVEE: Interpretable variant effect prediction from genomic foundation model embeddings

Scientific foundation models learn high-dimensional representations from diverse data modalities, yet what they encode and how to extract that knowledge remain open questions. Here we show that probing the internal representations of Evo 2, a 7-billion-parameter genomic foundation model, enables accurate and interpretable genetic variant effect prediction. We introduce a covariance-based probe that captures second-order structure from Evo 2 sequence embeddings to predict variant pathogenicity across variant types and functional consequences, matching or exceeding specialized predictors within their domains. To ground these predictions in known biological mechanisms, we train a complementary panel of probes on existing annotations to detect which genomic properties are disrupted by a variant. This categorized evidence is then integrated with each variants genomic context through a language model to generate variant-specific mechanistic hypotheses. Our pathogenicity predictions correlate with experimental measures of variant function, clinical penetrance, and biobank disease associations while the mechanistic hypotheses are consistent with expert reviews, known mechanism classes, and downstream molecular readouts. We release pathogenicity scores, disruption profiles, and contextualized interpretations for 4.2 million variants from the ClinVar database as an open resource through the Evo Variant Effect Explorer (EVEE). More broadly, this structured probing approach offers a general framework for interrogating foundation models across scientific disciplines and grounding their outputs in existing domain concepts.

genomics↗

bioRxiv2025

Efficient Protein Engineering via Integrated Language Models and Bayesian Optimization

This study investigates the application of advanced predictive models to reduce the cost and effort associated with protein engineering campaigns. We explore the use of protein language models (PLMs), a variant of large language models (LLMs), to predict functional performance from protein sequences. A common challenge in this domain is the scarcity of functional data. To address this, we examine zero-shot and few-shot learning methods. Another challenge is efficiently searching the vast fitness landscape for superior protein variants. We evaluate search methods, such as Bayesian optimization, to tackle this problem. The proposed methods are evaluated against a benchmark of 34 protein datasets containing sequences and their quantified functional values. Our findings demonstrate the potential of these advanced predictive models to streamline and accelerate the protein engineering process.

bioengineering↗

bioRxiv2025

Nuclear envelope budding is a non-canonical mechanism to export large transcripts in muscle cells

In recent years, nuclear envelope budding (NEB) has emerged as an alternative route for nuclear export of viral particles that are too large to pass through the nuclear pore complex. Yet, the significance of this unconventional export pathway for large endogenous cargoes in mammalian cells has remained largely unexplored. Here, we use a combination of electron and fluorescence microscopy to demonstrate that NEB events occur following myoblast differentiation into myotubes and concomitant with the expression of extremely long muscle-specific transcripts. We show that NE buds are derived from the inner nuclear membrane, contain internal vesicles, and are specifically enriched with long sarcomeric transcripts. We identify a role for the protein UIF in regulating mRNA cargo targeting into NE buds and show that this pathway requires the ESCRT-III membrane remodeling machinery. Our findings uncover a non-canonical pathway for large transcript nuclear export in muscle cells and provide insight into its mechanism.

cell biology↗