Chapter 10 TME Modeling

Previous studies have shown that the tumour microenvironment is a complex ecosystem. No single cell or gene is sufficient to influence the phenotype. Therefore, machine learning models of the tumour microenvironment or models of tumour microenvironment typing are used to predict tumour phenotypes and treatment response. In the last section, we present common considerations and scenarios for constructing tumour microenvironment models.

10.1 Loading packages

library(IOBR)

10.2 Data prepare

Using data from IMvigor210, we demonstrate two common scenarios for building models of the tumour microenvironment: predicting survival and predicting treatment response (BOR, RECIEST 1.1).

data("imvigor210_sig", package = "IOBR")
data("imvigor210_pdata", package = "IOBR")

10.3 Input data (overall survival) prepare

pdata_prog <- imvigor210_pdata %>% 
  dplyr::select(ID, OS_days, OS_status) %>%
  mutate(OS_days = as.numeric(.$OS_days)) %>% 
  mutate(OS_status = as.numeric(.$OS_status))

head(pdata_prog)

## # A tibble: 6 × 3
##   ID              OS_days OS_status
##   <chr>             <dbl>     <dbl>
## 1 SAM00b9e5c52da9    57.2         1
## 2 SAM0257bbbbd388   469.          1
## 3 SAM025b45c27e05   263.          1
## 4 SAM032c642382a7    74.9         1
## 5 SAM04c589eb3fb3    20.7         0
## 6 SAM0571f17f4045   136.          1

10.4 Constructing survival prediction models

prognostic_result <- PrognosticModel(x           = imvigor210_sig, 
                                     y           = pdata_prog, 
                                     scale       = T, 
                                     seed        = 123456, 
                                     train_ratio = 0.7, 
                                     nfold       = 8,
                                     plot        = TRUE)

## NULL
## NULL
## NULL

10.5 Input data (Response) prepare

pdata_group <- imvigor210_pdata[!imvigor210_pdata$BOR_binary=="NA",c("ID","BOR_binary")]
pdata_group$BOR_binary <- ifelse(pdata_group$BOR_binary == "R", 1, 0)
head(pdata_group)

## # A tibble: 6 × 2
##   ID              BOR_binary
##   <chr>                <dbl>
## 1 SAM0257bbbbd388          0
## 2 SAM025b45c27e05          0
## 3 SAM032c642382a7          0
## 4 SAM0571f17f4045          0
## 5 SAM065890737112          1
## 6 SAM0684af734db1          1

10.6 Constructing prediction models for response

binom_res <- BinomialModel(x           = imvigor210_sig, 
                           y           = pdata_group, 
                           seed        = 123456, 
                           scale       = TRUE, 
                           train_ratio = 0.7, 
                           nfold       = 8, 
                           plot        = T)

## NULL
## NULL
## NULL

## NULL

## NULL

10.7 References

Cristescu, R., Lee, J., Nebozhyn, M. et al. Molecular analysis of gastric cancer identifies subtypes associated with distinct clinical outcomes. Nat Med 21, 449–456 (2015). https://doi.org/10.1038/nm.3850

CIBERSORT; Newman, A. M., Liu, C. L., Green, M. R., Gentles, A. J., Feng, W., Xu, Y., … Alizadeh, A. A. (2015). Robust enumeration of cell subsets from tissue expression profiles. Nature Methods, 12(5), 453–457. https://doi.org/10.1038/nmeth.3337;

Seurat: Hao and Hao et al. Integrated analysis of multimodal single-cell data. Cell (2021)