Avi Srivastava, Ph.D.

Most cancer cells exposed to targeted therapy die, but a small population survives in a non-genetic, drug-tolerant persister state and seeds eventual relapse. The molecular basis of this state is poorly understood at a critical level of detail: which transcripts are being produced, in which cells, and at what moments during the response to therapy.

Standard single-cell RNA-seq cannot see this because it counts genes, not isoforms. Persister cells may differ from their drug-sensitive neighbors not only in which genes they express, but in which isoform of each gene, with potential consequences for protein localization, stability, and downstream signaling.

Using BenchDrop-seq and Bagpiper, we are mapping isoform-level transcriptional state across drug-tolerant persister populations. The goal is to identify splicing switches that are causal for persistence and therefore druggable.

Epstein-Barr virus has been firmly linked to multiple sclerosis, but the mechanism by which a near-universal latent infection produces autoimmune disease in a small subset of carriers remains unclear. Splicing dysregulation is increasingly implicated, but standard tools cannot resolve which immune cell populations are splicing-rewired or how those isoform changes connect to the disease state.

In collaboration with the Lieberman Lab, we are applying BenchDrop-seq to immune cell populations from EBV-positive donors with and without MS, with the goal of identifying splicing signatures of dysregulated immune cell fate that gene-level analysis structurally cannot resolve.

BenchDrop-seq is a microfluidics-free single-cell platform that combines PIP-seq capture with Oxford Nanopore long-read sequencing on a standard lab bench, requiring no dedicated instrument. Bagpiper is the computational pipeline that turns raw Nanopore reads into a per-cell, per-isoform count matrix. It performs spacer-anchored barcode recovery, junction-aware read assignment, and expectation-maximization-based resolution of multi-mapping reads. Bagpiper makes BenchDrop-seq quantitatively interpretable and is also applicable to long-read single-cell data generated on other platforms. Github: https://github.com/avisrilab/bagpiper

The lab’s leadership contributed to several widely used computational genomics platforms during graduate and postdoctoral training, which continue to be developed and maintained by their original groups:

– salmon, alevin, alevin-fry: bulk and single-cell RNA-seq quantification with explicit uncertainty propagation. Patro Lab, University of Maryland.
– Seurat: integrated multimodal single-cell analysis. Satija Lab, NYGC.
– Signac: single-cell chromatin state analysis. Satija and Stuart Labs.
– scCUT&Tag-pro: simultaneous chromatin modification and surface protein profiling. Satija Lab.

These tools are widely used across the single-cell genomics field.