Skip to content

Alexandra Bradu — VIPerturb-seq + the "GuEST-List" library

Speaker: Alexandra (Alex) Bradu, Associate Research Scientist — Satija Lab, New York Genome Center. Slot: SCGD26, Jun 12, 2026 · 2:45–3:05 PM EDT — "Genome-wide single-cell perturbation screens with VIPerturb-seq" Method: VIPerturb-seq ("Very Important Perturbations") + the GuEST-List genome-wide guide-barcode library (Guides for Enrichment-based Screening of Targets; the transcript's "Guest List" is the spoken acronym) Recording: SCGD26 livestream — preprint, project site + protocols, and Addgene pooled library (#248188/#248189) available Status: Captured from session transcript

Thesis

Make genome-wide single-cell CRISPR screens routine and cheap. Published genome-wide screens needed >1,000 lanes of 10x. VIP Perturb-seq attacks cost on two axes: (1) raw throughput, and (2) phenotypic enrichment so you only deep-sequence the cells you care about.

How it works

Marries a genome-wide perturbation library (e.g. Dolcetto CRISPRi, Doench lab) with 10x Flex:

  • Throughput: Flex's newest chemistry adds a barcode to subsets of cells before GEM formation → pack multiple cells per droplet → genome-scale screen "in a single afternoon."
  • Enrichment: Flex works on fixed/permeabilized cells, so you can FACS-sort on intracellular protein / PTM / even RNA (not just surface markers) and only build libraries from enriched cells.
  • The probe problem + "Guest List": Flex needs a spiked-in probe per transcript; custom probes against every guide protospacer would cost six figures and aren't reusable. Instead, each guide is cloned with a synthetic barcode (modified CROP-seq design). The barcode is split into part-A × part-B (100 × 100 options → 10,000 constructs) so <500 probes read out the whole Dolcetto library. Recombination between protospacer and barcode is kept <2% by minimizing distance (<100 bp) and using three low-homology (<20 bp) guide scaffolds. The genome-wide library is "GuEST-List" (Guides for Enrichment-based Screening of Targets — every gene invited; only some become VIPs per experiment); its protospacers match the Dolcetto A CRISPRi library (57,050 guides, >18,000 genes), deposited on Addgene (#248188 puro; #248189 puro+eGFP).

Key results

  • Clean, unambiguous guide assignment (e.g. FDFT1: hundreds of UMIs for the assigned guide, single digits for the next) with robust target knockdown; >500 probes detect the whole Dolcetto library.
  • Throughput: a single 10x lane recovers a huge cell count vs. weeks of prior work; improved sensitivity for lowly expressed transcripts vs. 3′ capture.
  • Enrichment proof-of-concept (intracellular Vimentin): sort Vimentin-high K562s, input ~20k cells → clean enrichment of the Rag-Regulator/Ragulator (mTOR) complex as VIPs, at ~2% of the genome-wide cell count.
  • RNA-phenotype screens: integrated with Caleb LaReau's perf-seq (RNA flow-FISH isolation of rare types) — pilot CRISPR screen on a non-coding-RNA phenotype with clean positive-control enrichment (postdoc Aaron Cummings).

Q&A highlights

  • Enrichment threshold chosen per phenotype via arrayed FACS benchmarking + mixing experiments; multiple sort bins can be sample-indexed in one run.
  • Split-barcode capture generalizes beyond guides — ORF libraries, lineage cassettes, combinatorial screens. 10x is seeking early-access customers for the commercial version.

Connections to the corpus

Sources

Note: the project site frames VIPerturb-seq as openly shared reagents/protocols and reports a ~50-fold throughput gain via combinatorial indexing; it does not mention a 10x commercial product (the "commercial version" claim from the transcript is unconfirmed).