Repressing the immune resistance program

To identify drugs that could potentially be selectively lethal to immune resistant cancer cells we:

1. Computed the overall expression of the immune resistance program across various human cancer cell lines, spanning XX cancer types.

r$res<-cmb.res.scores(r = r,res.sig = res.sig,bulk.flag = T)

2. Used a logistic regression mixed-effects model that controls for the different cancer types, to identify drugs that are more toxic to immune resistant cancer cell lines. We used the IC50 values of 131 drugs across 639 human cancer cell lines as reported in (Garnett et al., 2012).

M1 <- with(r0, glmer (y ~ state + (1 | types), family = binomial(link="logit"),
                          control=glmerControl(optimizer="Nelder_Mead",optCtrl=list(maxfun=20000))))

Where types denotes the cancer type, y denotes if the cell line is sensitive to the drug or not, and state denotes if the cell line is immune resistant or not, according to the program.

The code is provided in ImmRes6_pancanDrug

find.sensitizing.drugs()
[1] "Garnett et al. 2012, top 5 immune sensitizing drugs:"
                    Z  Estimate     Pr(>|z|)
Vorinostat   5.981066 1.5646287 2.089124e-06
FH535        4.556426 1.3720329 5.553972e-05
PD.0332991   4.294721 1.0937556 1.014633e-04

[1] "Gong et al. 2017, CDK4/6i is more effective in immune resistant cell lines"
[1] "P = 3.96e-05"

Among the topmost scoring drugs was a CDK4/6-inhibitor (PD.0332991, also known as palbociclib, ranked 3rd, P = 1.01*10-4, mixed-effects).