Tutorials#
Tutorials are hands-on notebooks. The core workflow teaches the AlphaGenome 2 API in order: choose a condition, select a genomic region, predict, and score. Each topic tutorial then covers one capability. For end-to-end biological analyses, see Vignettes.
Core workflow#
Search supported_conditions, check which output types are observed or
imputed for a condition, and see when to use expression conditioning instead.
Construct and resize Interval and Variant inputs around coordinates,
GENCODE genes, or variants.
Predict bulk and pseudobulk tracks for a genomic interval, and bulk tracks for a custom DNA sequence.
Predict REF and ALT bulk and single-cell outputs for a genetic variant and
plot them together.
Quantify variant effects with recommended and custom variant scorers, and
flatten the results with tidy_scores.
Topics#
Compose multi-panel plots across output types: bulk coverage, single-cell gene expression, chromatin accessibility, histone and TF ChIP-seq, and contact maps.
Condition predictions and scores on your own AnnData single-cell gene
expression counts.
Check which tracks are observed or imputed for a condition, filter by
imputation_mode, and label imputed tracks in plots.
Predict, plot, and score splice site, splice site usage, and splice junction changes for a variant.
Predict and compare combined multi-variant haplotype effects against individual single-variant predictions.
Run in silico mutagenesis (ISM) across an interval to find the bases that drive a prediction.
Load mouse gene annotations and make predictions for the mouse genome.
AlphaGenome 1#
AlphaGenome 1 is the original model described in the AlphaGenome paper [Avsec et al., 2026]. See Getting started for when to use it and Migrating from AlphaGenome 1 for how it differs from AlphaGenome 2.
Browse AlphaGenome 1 metadata, then predict and score a variant across all observed tracks.
Compute percent spliced in (PSI5 and PSI3) values from predicted splice junction counts.