search

drjingma / research-plan

summary of all ongoing and planned research topics
0 stars 0 forks source link

Literature review #26

Open drjingma opened 6 days ago

drjingma commented 6 days ago

We would like to understand what's the new advances in analysis for Perturb-seq data.

Data type

  1. single cell CRISPR screens (Perturb-seq) with low MOI: one perturbation per cell
  2. Perturb-seq with high MOI: multiple perturbations per cell
drjingma commented 6 days ago

Background Current methods study differential expression where the control/treatment groups correspond to cells with non-target perturbations vs cells with target perturbations. Several analytical challenges were identified: sparsity in perturbations and sparsity in gene expressions, confounding due to technical factors, and model misspecification.

Knowledge gaps

  1. DE methods do not model gene co-expression.
  2. DE methods cannot distinguish direct from indirect effects in the sense that perturbations can be associated with their direct targets or with targets further downstream.
  3. some non-target guide RNAs may have off-targeting effects.
  4. Some targeting gRNAs are ineffective.

The first two issues can be potentially solved by using multivariate methods, such as factor analysis or network-based methods.