AIMS-ESI release resources
==========================

AIMS-ESI: An Adaptive Multiview Interaction Framework for
Enzyme–Substrate Compatibility Prediction and Biocatalytic Discovery

Packages
--------

1. AIMS-ESI-source-code.zip
   Model implementation, frozen configurations, preparation/training/
   evaluation scripts, environment declarations, and reproducibility
   manifests.

2. AIMS-ESI-benchmark-resources.zip
   Benchmark input manifests, fixed split artifacts, released prediction
   outputs, metrics, and non-checkpoint reproducibility records.

Quick start
-----------

Create the verified environment:

  conda env create -f environment.yml
  conda activate aims-esi

Validate the release and reproduce the AIMS-only stacking outputs:

  python scripts/verify_release.py
  python scripts/stack_aims_only.py --split all

Canonical dataset SHA-256
------------------------

e60d682a095084fadb246791203f2bd670b2a82f83ef2f00ada776190f31efb6

Important reproducibility rule
------------------------------

Do not select checkpoints, ensemble weights, or decision thresholds using
test labels. Comparator-model predictions must not be mixed into an output
reported as AIMS-ESI.

Website: http://bliulab.net/AIMS-ESI