Data and models for: An End-To-End Computational Framework for 'Record-seq' Transcriptional Recording Data

This repository contains the trained neural models and associated datasets described in the manuscript "An End-To-End Computational Framework for 'Record-seq' Transcriptional Recording Data" (Hugi et al., 2026). The provided files allow for the reproduction of the acquisition modeling results. Contents Trained Neural Models: Model weights and architectures for the sequence-based neural networks (CNN-BiLSTM). These models are designed to predict base-resolution acquisition "acquirability" from primary DNA sequences and genomic annotations. Training and Validation Data: Normalized spacer coverage profiles derived from multiple published E. coli Record-seq experiments. These profiles serve as the ground-truth targets for model training. Genomic Features: Preprocessed input features including one-hot encoded DNA sequences and annotation channels (gene bodies, promoters, terminators, and orientation). Related Code Manuscript main repository Spacer acquisition modelling repository