Core-code

• Core codes when studying in Guangdong Provincial People's Hospital, Guangzhou.
• Personal use & ready to share.
• Most codes are adjusted from tutorials of related packages/softwares/articles.
• Reference of each package/software/method is omitted.

Contents

0.Notes (in Chinese)

• Notes during studying and code writing.

1.Disease definition for UKB

• R
• ICD10 & OPSC4
• Disease contains IHD, HF...

2.MICE to impute missings in covariates

• R
• 5 times of imputation
• Rubin rules

3.Cox model to get HR

• R/Python
• In most cases, plasma proteins are risk factors
• Covariates adjustment

4.Machine learning model for prediction

• Python
• kmeans, LGBM, RF, SVM, XGBoost, DT, LR...
• SMOTE/Downsampling to deal with imbalance problems
• Bayesian-optimization
• Accuracy, precision, recall...
• ROC AUC curve & PR curve
• Cross-validation (k- fold, External validation...)

5.KM analysis

• Python
• Youden Index to determine cutoffs for each protein (biomarker effect)
• Numbers at risk & numbers at event

6.MR analysis

• Python/R
• Gwaslab to wash data into vcf framat in GR37/hg19
• Series of IEU packages to conduct Two-sample MR analysis locally

7.Colocalization analysis

• R
• Gwasglue to match vcf files in given chrompos
• Conc for colocalization analysis

8.GWAS & PRS

• sh/R
• Plink2 to perform GWAS analysis
• PRSice-2, LDpred2, and lassosum for PRS calculation

9.PLACO

• R
• Pleiotropic effect of a genetic variant on two traits
• Reference: https://github.com/RayDebashree/PLACO