The RAGE Project

class: center, middle, inverse, title-slide

.title[
# The RAGE Project
]
.author[
### John Thompson
]
.date[
### 29th March 2023
]

---

#RAGE

**R**egression **A**fter **G**ene **E**xpression

##Aim

- To explore ways of selecting features from a gene expression study for prediction of a patient characteristic.

- To investigate the pros and cons of features based on the principal components of the gene expression.

---
# Data Source

The gene expression data come from,

Sayed N, Huang Y, Nguyen K, Krejciova-Rajaniemi Z et al.  
**An inflammatory aging clock (iAge) based on deep learning tracks multimorbidity, immunosenescence, frailty and cardiovascular aging.**  
Nat Aging 2021 Jul;1:598-615.

GEO archive as GSE168753. File **GSE168753_processed_data.csv.gz** from https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE168753.

---
#Study Design

- 312 individuals 
 
- Profiled using Illumina HumanHT-12 V4.0 expression beadchip 
 
- Processed data has expression for 5000 genes 
 
- Subject characteristics include, age, sex, BMI

---
#Project Plan
 
* Use gene level expression to predict BMI 
 
* Develop the analysis on a sub-sample of 1000 genes 
 
* Use 200 subjects for training the models and 112 subjects for validation 
 
* Predict BMI using Linear Regression (LReg) 
 
* Compare 3 methods 
 - Rank Probes .. Select the M best probes for LReg 
 - PCA of Correlations .. Select first M PCs for LReg 
 - Rank PCs .. Select the M best PCs for LReg 
 - PCA of Covariances .. Select first M PCs for LReg 
 - Rank PCs .. Select the M best PCs for LReg 
 
* Once code is working, use same workflow on all 5000 genes and 312 subjects

---
#Loss Function

Compare models using the Least Squares Loss

$$
  \frac{1}{N} \sum_{i=1}^N (y_i -\hat{y}_i)^2 
$$

N is the number of subjects,  
`$y_i$` is the BMI  
`$\hat{y}_i$` is the predicted BMI under whatever model is being used

- Measures the average prediction error  
- Equivalent to -log-Likelihood of the normal distribution used in linear regression

---
#Documentation

- Commented Scripts, Diary, Dashboard, Reports on all intermediate stages 
 
- Slide presentation on the final analysis 
 
- Draft journal article (PLoS) 
 
- Website describing the project