porecast

Forecast electrophysiology from single-cell ion-channel expression.

porecast maps the ion-channel pore composition of any cell type (from a single-cell RNA-seq dataset) to a forecast of its electrical behaviour. It is the method companion to the CATION ion-channel cell atlas — the ion-channel analogue of hormone2cell, with one key difference: ion channels are the only gene family where expression maps quantitatively to a measurable biophysical output (the action potential, V(t)), so porecast can predict function, not just describe expression.

Planned API (three tiers)

import porecast as pc

# tier 1 — score the ion-channel fingerprint of every cell type
fp = pc.fingerprint(adata)

# tier 2 — classify cell types as electrically excitable vs non-excitable
pred = pc.excitability(adata)          # validated against causal CRISPR-perturbation ephys

# tier 3 — infer a Hodgkin–Huxley model and forecast the action potential
hh = pc.model(cell_type="L5 ET pyramidal")
hh.spike(current_injection=40)         # -> predicted V(t)

porecast ships with the curated IUPHAR/BPS channelome (320 genes, 55 families), a model card, and trained weights.

Status

0.0.1 — name reservation. This release reserves the package name on PyPI. The full implementation lands with the CATION atlas publication. Watch this space; see the CATION atlas.

Installation

pip install porecast

License

MIT.