Multi-Omics Integration and Single Cell RNA-Seq (Perturb-Seq)

Capturing biology at scale: integrating diverse data layers

Multi-Omics Integration

Unifying diverse data layers into coherent variant–pathway maps.

Background

Variants exert effects at multiple levels: transcription, translation, protein signaling, and functional phenotypes. Individually, datasets like genomics, transcriptomics, proteomics, and CRISPR screens offer only partial views. The challenge is to integrate across omics layers to capture emergent biology.

Approach

• Built pipelines to harmonize genomics + transcriptomics + CRISPR + proteomics.

• Applied unsupervised ML (NMF, clustering) to uncover latent structures.

• Used graph neural networks (GNNs) to represent variant–pathway relationships across omics.

• Constructed multi-omic embeddings that align diverse datasets into unified maps.

Outcome

• Identified cross-validated variant clusters with pathway-level impacts.

• Improved robustness of variant annotation compared to single-omics pipelines.

• Enabled downstream drug response predictions informed by multiple data types.

Single-Cell RNA-seq & Perturb-seq

Zooming in: variant effects at single-cell resolution

Background

Bulk assays obscure cellular heterogeneity. Single-cell RNA-seq (scRNA-seq) and especially Perturb-seq (scRNA-seq combined with CRISPR perturbations) allow mapping variant effects on gene expression programs at single-cell resolution.

Approach

• Developed pipelines for scRNA-seq and Perturb-seq preprocessing, clustering, and visualization.

• Applied gene set scoring, differential expression, and network analysis to interpret CRISPR perturbations.

• Built a reproducible pipeline for analyzing Perturb-seq datasets.

Case Study (in progress)

• Applying Perturb-seq analysis to TNBC organoids.

• Identifying variant-driven clusters that disrupt oncogenic pathways.

• Validating computational predictions with experimental CRISPR knockdowns.