R error on train data using caret

Question

When I try to train some data on a simple lm in caret it return this:

> rdati<-read.csv("cutDati.csv")
> 
> rdati[]<-lapply(rdati, factor)
> 
> dataM=as.data.frame(rdati[])
> dataM=dataM[1:100,]
> 
> data=na.omit(dataM)
> trainIndex <- createDataPartition(data$d90, p = .8,
+                                   list = FALSE,
+                                   times = 1)
Warning messages:
1: In createDataPartition(data$d90, p = 0.8, list = FALSE, times = 1) :
  Some classes have no records ( 4, 6, 7, 9, 10, 11, 14, 15, 16, 17, 20, 21, 22, 23, 25 ) and these will be ignored
2: In createDataPartition(data$d90, p = 0.8, list = FALSE, times = 1) :
  Some classes have a single record ( 1 ) and these will be selected for the sample
> training <- data[ trainIndex,]
> testing  <- data[-trainIndex,]
> 
> fit <- train(d90 ~ ., data = training,
+                  method = "svmLinear",
+                  verbose = FALSE)
Error in `contrasts<-`(`*tmp*`, value = contr.funs[1 + isOF[nn]]) : 
  contrasts can be applied only to factors with 2 or more levels

d90 contain factor from -1 to 27 without 0.

Can somebody help me? :)

Thanks, Marco

Edit:

add some infromation

> str(training$d90)
 Factor w/ 25 levels "-1","1","2","3",..: 24 2 14 4 13 6 4 4 14 6 ...
> unique(training$d90)
 [1] 26 1  13 3  12 5  27 8  -1 2 
Levels: -1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 20 21 22 23 25 26 27

I discover that if i restrict the area it works.

training <- data[ trainIndex,200:300]
testing  <- data[-trainIndex,200:300]

fit <- train(d90 ~ ., data = training,
                 method = "svmLinear",
                 verbose = FALSE)

I'm thinking about the problem is on the data, but what problem could be? How i can check?

I found the problem but i don't understand how i can resolve it

I try some combination and i saw that the bug is in column 205

> unique(training[,205])
Error in `[.data.frame`(training, , 205) : 
undefined columns selected

Can you help me?