A histogram-based neural network with LLM gene embeddings predicts full expression distributions for unseen genetic perturbations, capturing variance, skewness, and kurtosis better than mean-only baselines at lower training cost.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
q-bio.GN 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Modeling Gene Expression Distributional Shifts for Unseen Genetic Perturbations
A histogram-based neural network with LLM gene embeddings predicts full expression distributions for unseen genetic perturbations, capturing variance, skewness, and kurtosis better than mean-only baselines at lower training cost.