Perturbation modeling

[sunnyosun]

3-level concepts

1. Perturbation target (pert_target)

Targets can be a gene, a protein, or a pathway depending on the perturbagen and condition.

When applying genetic pertubagens, for example a CRISPRi experiment, registering perturbation targets is sufficient if we just want to model the target genes and doesn't care about which guide RNAs are used.

2. Perturbagen (pert_name)

perturbation_type: compound, biologic, genetic, physical

1. Compound (small molecules, chemical entities): chemical properties, ChemBL IDs, etc.
   • Small molecule drugs
     • Imatinib (Gleevec) - cancer drug targeting BCR-ABL
     • Rapamycin - immunosuppressant targeting mTOR
     • Tamoxifen - selective estrogen receptor modulator
   • Chemical inhibitors
     • DMSO (dimethyl sulfoxide) - common solvent and cryoprotectant
     • MG132 - proteasome inhibitor
     • Actinomycin D - transcription inhibitor
   • Natural products
     • Cyclosporine - immunosuppressant from fungi
     • Paclitaxel (Taxol) - chemotherapy from yew tree bark
     • Tetrodotoxin - neurotoxin from pufferfish
   • Metabolites
     • Glucose - sugar molecule
     • ATP (adenosine triphosphate) - energy carrier
     • Amino acids like glutamine
2. Biologic (large molecules): protein sequences, molecular weight, etc.
   • Proteins
   • Peptides
   • Antibodies
   • Enzymes
   • Growth factors
   • Viral infections (they're biological agents causing direct perturbation, different from viral vector which belongs to genetic perturbagen)
3. GeneticPerturbagen (genetic tools like CRISPR, shRNA): system type, sequences, target scores
   • CRISPR-based tools
     • CRISPR-Cas9 knockout targeting TP53 gene
     • CRISPRi (interference) targeting MYC expression
     • CRISPRa (activation) enhancing IL2 expression
   • RNA interference
     • shRNA targeting BRCA1
     • siRNA against EGFR
     • miRNA mimics or inhibitors
   • Overexpression systems
     • Inducible GFP expression construct
     • Constitutive KRAS mutant expression
     • Tet-ON/OFF systems
   • Other gene editing tools
     • TALENs targeting BCL2
     • Zinc finger nucleases
     • Base editors modifying specific nucleotides
     • Viral vector
4. PhysicalCondition (for things like heat shock, UV, mechanical stress): intensity, duration, units of measurement, etc. (The reason to use PhysicalCondition rather than PhysicalPerturbagen is that, "perturbagen" typically implies a molecular entity. Physical factors are usually described as conditions or parameters in experimental biology.)
   • Heat shock
   • UV radiation
   • Mechanical stress
   • Ultrasound
   • Electric fields

Modeling controls

• For Compound

  • DMSO vehicle control
  • Water/saline control
  • Mock treatment
  • Untreated control

• For GeneticPerturbation

  • Non-targeting guide RNA (scrambled sequence)
  • Empty vector control
  • Safe harbor targeting (like AAVS1)
  • Luciferase control guide

3. Perturbation

pert_dose, pert_time

E.g. A compound is applied at xnM (dose) for yh (duration). This is not commonly registered in a registry and is also complicate to validate.

[Zethson]

Generally good! How would you model combinations of two types of perturbagens now?