Pia Rautenstrauch

The paper argues against using Silhouette metrics for evaluating single-cell integration methods due to their unsuitability in handling batch effects and proposes alternative strategies.

Liam integrates multimodal single-cell data by learning a joint low-dimensional representation, outperforming existing methods on Multiome and CITE-seq datasets.

The paper performs a rare variant association study of 30 blood biomarkers in 191,971 UK Biobank individuals, identifying 193 gene-biomarker associations using functionally informed kernel-based tests.

A NeurIPS 2021 benchmarking competition evaluated over 2600 submissions from 280 competitors across three multimodal single-cell integration tasks using a purpose-built benchmarking dataset and cloud evaluation pipeline.