How to change the color legend in fine granularity using ggplot2

Question

I am very new to ggplot2 and trying to create 9 geom tile plots using this melted dataset and the below code

cm <- read.table("http://dl.dropboxusercontent.com/u/7147744/cm_melted.log", header=T, sep=" ")
xlab = expression(italic(lb)~(Normalised))
ylab = expression(italic(dti)~(Normalised))

ggplot(cm, aes(x=LB, y=DTI)) +
  facet_wrap(~Method, ncol=3) +
  geom_tile(aes(fill=CM), colour="white") +
  scale_fill_gradient(low="white", high="steelblue") +   
  theme_bw() +
  coord_equal() +
  xlab(xlab) +
  ylab(ylab) +
  ggtitle("Comparison of different methods based on composite metric")

I would like to change the colors of the tiles and the legend in more fine granularity. For example if the minimum value of CM is 0 and maximum is 1, then the legend and tiles should be made up from 11 different colors for 0.1 step size.

Another concern is how can I turn the title color for the NA values in `CM' into white so that the main focus is only with the lower triangle.

And the last thing is can I use black and white pattern fills for these plots? Or any other suitable suggestions for print quality plots in bw.

Many thanks.

Update1:

while using geom_tile(aes(fill=factor(value), colour="white")) + I can get the below plot as it is color coded based on every distinct value of CM

Answer 1

Here's what you can start with. You said that the scale should be from 0 to 1, but in your example maximum is about 0.6, so I took that into consideration:

p <- ggplot(cm, aes(x=LB, y=DTI)) +
       facet_wrap(~Method, ncol=3) +
       geom_tile(aes(fill=CM), colour="white") +
       theme_bw() +
       coord_equal() +
       xlab(xlab) +
       ylab(ylab) +
       ggtitle("Comparison of different methods based on composite metric")

# n equally placed breaks for n colours
n_breaks <- 10
br <- c(0, max(na.omit(cm$CM)))
split_interval <- function(v, n) seq(from=v[1], to=v[2], length.out=n)
p + scale_fill_gradientn(colours = rainbow(n_breaks), 
                         na.value = "white", 
                         breaks = split_interval(br, n_breaks))

Play a bit with breaks and number of colours to get the most suitable picture. Check the available palettes, the default hue should probably be more appropriate.

Answer 2

Thanks to @tonytonov for his great answer. I am also sharing what I was trying in the meanwhile and the output. [Note: modified the dataset cm to include method names instead of numbers from 1 to 9]

max_cm <- max(na.omit(cm$CM))
min_cm <- min(na.omit(cm$CM))

ggplot(cm, aes(x=LB, y=DTI)) +
  facet_wrap(~Method, ncol=3) +
  geom_tile(data= subset(m, !is.na(CM)), aes(fill=CM)) +
  scale_fill_gradientn(colours=c("darkgreen", "green", "greenyellow", "yellow", "red", "darkred"), breaks=seq(0, 1, 0.1), "CM", limit=c(min_cm, max_cm)) +
  theme_bw() +
  coord_equal() +
  xlab(xlab) +
  ylab(ylab) +
  ggtitle("Comparison of different methods based on composite metric")