Training School: Risk modeling in colorectal cancer

Identifying Biomarkers Through Translational Research for Prevention and Stratification of Colorectal Cancer (TRANSCOLONCAN)

Monday 21st

14:00-16:00 Basics of predictive modeling (Victor Moreno)

• Study design
• Analysis strategy
• Data preprocessing
• Over-fitting. Internal and external validation.
• Statistics for predictive modeling
• Assessment of accuracy
• Model calibration

16:30-18:30 Computer exercises (M.Henar Alonso & Victor Moreno)

• Simple model building (logistic regression)
• Assessment of predictive accuracy
• Bootstrap and cross validation with R

Tuesday 22nd

09:00-11:00 Predictive modeling tools and strategies (Victor Moreno)

• Logistic regression
• Elastic nets
• Support vector machines
• Trees and Random forests
• Boosting
• Neural networks
• Feature selection and predictor importance

11:30-13:00 Computer Exercises of analysis with R (M.Henar Alonso & Victor Moreno)

• Model building
• Model assessment
• Feature selection

13:00-14:00 Lunch

14:00-16:00 Genetic risk models (Victor Moreno & Robert Carreras)

• Overview of genome-wide association studies (GWAS) data and analysis
• Genetic risk scores
  • Building genetic scores
  • Models with genetic scores
• Mendelian randomization
• GWAS correlations

16:30-18:30 Computer exercises (Robert Carreras & Anna Diez)

• Capitalizing GWAS data to identify risk factors
  • Mendelian randomization (hands-on training; MR-base)
  • GWAS correlations (hands-on training; LD-Hub)

Wednesday 23rd

09:00-11:00 Integrating GWAS and gene-expression data (Robert Carreras)

• Transcriptome-wide association studies (TWAS) overview
• TWAS correlations
• Overview of mutational signatures on somatic tumor tissue

11:30-13:00 Examples of genetic risk models (Sergi Castellvi)

• Paper review of paradigmatic examples
• Reporting guidelines STARD, REMARK
• Public health interpretation of risk models

13:00 Lunch