Kipoi - Seminar

The monthly virtual seminar series is designed as a platform for interested Kipoi users and developers and will host talks on the applications of deep learning on biological data. The seminar is held on every first Wednesday of the month at 5:30 p.m. - 6:30 p.m. CET. We are also happy to share the recordings of the seminar on YouTube.

How to take part

The Virtual Seminar Series takes place via Zoom. To take part in the seminar, you can register for the online Zoom conference. Your personal join link will be valid for all upcoming lectures of the series.

Register

How to apply as a speaker

The seminar is a great opportunity to present your recent work to a large international audience. If you want to apply as a speaker, please use the contact in the registration confirmation email.

Next seminar

Title: AlphaGenome: advancing regulatory variant effect prediction with a unified DNA sequence model

4 February 2026 5:30 p.m. - 6:30 p.m. Central European Time

Speaker: Jun Cheng, DeepMind

Abstract:

Deep learning models that predict functional genomic measurements from DNA sequence are powerful tools for deciphering the genetic regulatory code. Existing methods trade off between input sequence length, prediction resolution and modality coverage, thereby limiting their scope and performance. We present AlphaGenome, a unified framework that takes a 1 megabase (Mb) DNA sequence as input to predict thousands of functional genomic tracks at up to single base-pair resolution across diverse modalities. These outputs span 5,930 human and 1,128 mouse tracks, including gene expression, chromatin accessibility, histone modifications, DNA contact maps, and splice junction coordinates and strength. Trained on human and mouse genomes, AlphaGenome matches or exceeds the strongest available external models on 24 out of 26 evaluations on variant effect prediction and 22 of 24 genome track prediction tasks. This strong performance extends across multiple regulatory axes. For gene expression, AlphaGenome significantly improves the prediction of eQTL effect direction, recovering over twice as many GTEx fine-mapped eQTLs (41%) as the previous leading model (19%) at a 90% sign accuracy threshold. In splicing, a novel composite scorer integrating splice junction predictions achieves state-of-the-art performance on 6 out of 7 splicing-specific benchmarks, including improved classification of pathogenic versus benign ClinVar variants. For chromatin regulation, the model also consistently outperforms specialized models in predicting effect sizes and distinguishing causal from non-causal chromatin accessibility QTLs across diverse human ancestries. AlphaGenome's ability to simultaneously score variant effects across all modalities accurately recapitulates the complex mechanisms of clinically-relevant variants, as demonstrated near the TAL1 oncogene. For known gain-of-function mutations, the model correctly predicts the upregulation of TAL1 expression, the concurrent increase in local activating histone marks, and the de novo creation of a MYB transcription factor binding motif, mirroring experimentally validated oncogenic mechanisms. To facilitate broader use, we provide tools for making genome track and variant effect predictions from sequence.

Upcoming speakers

Previous speakers

The scientific committee