ROSMAP-preprocess.Rmd

---
title: "Pre-Process ROSMAP Data"
description: |
 This script removes low-expressed ROSMAP genes and adjusts by age, sex, and post-mortem interval (PMI).
output:
  distill::distill_article:
    toc: true
---

```{r setup, include = FALSE}
knitr::opts_chunk$set(eval = FALSE)
```

# Dependencies

Load requisite packages.

```{r load-packages}

library(data.table)

```


# Load Data

Load `.Rdata` object saved by prior script.

```{r load-data}

# load data
load("../Data/ROSMAP-24.Rdata")
rosmap_fpkm = rosmap$fpkm
cov = rosmap$cov

```


# Filter Genes

Remove low-expressed genes.

```{r filter-genes}

# remove low expressed genes
gene_sel = rowMeans(rosmap_fpkm) > 0.1
rosmap_fpkm = rosmap_fpkm[gene_sel, ]

```


# Adjust and Transform

Adjust by age, sex, and post-mortem interval (PMI), then log-transform data.

```{r adjust-covariates}

# create empty matrix to populate
adj_exp = matrix(NA, ncol = ncol(rosmap_fpkm), nrow = nrow(rosmap_fpkm))
colnames(adj_exp) = cov$projid
rownames(adj_exp) = rownames(rosmap_fpkm)

# adjust by age, sex, and PMI
msex = cov$msex
age_at_death = cov$age_at_death
pmi = cov$pmi

# impute missing values with mean
pmi[is.na(pmi)] = mean(pmi, na.rm = T)

```

Log-transform data.

```{r log-transform}

# log2(fpkm + 0.0001) adjusted by age, sex, and PMI
for(gene in rownames(rosmap_fpkm)){
  exp = log2(as.numeric(rosmap_fpkm[gene,]) + 0.1^4)
  lm_fit = lm(exp ~ msex + age_at_death + pmi)
  adj_exp[gene, ] = lm_fit$residuals
}

```


# Save Data

```{r save-data}

# save data
expSet = list(mData = adj_exp,
              cov = cov)
save(expSet, file = "../Data/ROSMAP-24-adj-low.expr.genes.removed.Rdata")

```