Dose response curves
- per compound
- per cell line
- for combinations (provide training set for learning algorithm)
Molecules
- molecular weight
- solubility
- chemical fragments
- shape
- QSAR
Genomic
- CNVs (copy number variations)
- parts of genome are {duplicated|lost}
- changes in expression levels -> gene dosage effects
- most important for gene loss
- SNPs - structural, splice sites, regulation
Methylation
- histones -> expression
Epistatic interactions
- A or B by itself not effective, need to stop both
- Pathway C might regulate/interact with A/B
Network data
- pathways (priority to pathways involved in cancer?)
- cell growth
- genome stability
- loss of contact inhibition
- metastasis
- PPi
- co-expression data (quantitative relationships b/w proteins and genes)
Feature -> ML -> Combination scores
Round 1 DeadLine has been changed to Dec 3.(See DeadLine File for more details)
- Infer Drug Synergy without experimental synergy score.
- Making drug predictions based on prior knowledge.
- We are allowed to use ** molecular data for the cell lines, cell response data for all respective mono-therapies, chemical information and putative targets of the compounds**
- Require 740 drug combinations without overlapping with Subchallenge 1.
-- 4a. A leaderboard set (370 combinations) and a final validation set (370 combinations) are required.
-- 4b. A full synergy-prediction-matrix (Score either 1 or 0)
-- 4c. A full synergy-confidence-matrix (Score ranging from 0 to 1) - The above data will be used to score accuracy of predictions.
- Justify our stratifications and synergy predictions, by giving an algorithm and rationally translatable as biomarkers.
- Boris (Hyginn)
- Emma (ehsueh)
- Julian (thejmazz)
- Nathan (njia95)
- Pruthvi (Pruthv1)
- Ashley (AshleyWWW)
- Jack (c5chenpe)
- Bhawan (B-1P)
- Targets: Nathan, Jack
- Data Preprocessing: Ashley, Fred
- Data Reduction: Boris
- Cross Validation: Chris, Julian
- Drug Similarity: Bhawan, Zach, Ricardo, Emma
- Cell Line Similarity: Pruthvi, Jenny
- Scoring:
- Submission: Jack, Nathan
- Documentation: Julian
- QA: