First commit of code and notebook

.gitignore

@@ -0,0 +1,5 @@
+.Rproj.user
+.Rhistory
+.RData
+.Ruserdata
+packrat/lib*/

immunotherapy notebook.Rmd

@@ -0,0 +1,71 @@
+---
+title: "Immunotherapy Notebook"
+output:
+  html_document:
+    df_print: paged
+---
+
+## Introduction
+
+Inspired by the blog post at https://blogs.rstudio.com/tensorflow/posts/2018-01-29-dl-for-cancer-immunotherapy/
+
+## Packages
+
+Packages for sequence logos and peptides
+
+```r
+devtools::install_github("omarwagih/ggseqlogo")
+devtools::install_github("leonjessen/PepTools")
+```
+
+## Analysis
+
+```{r}
+library(keras)
+library(tidyverse)
+library(PepTools)
+```
+
+Download and cache the data
+
+```{r}
+pep_file <- get_file(
+  "ran_peps_netMHCpan40_predicted_A0201_reduced_cleaned_balanced.tsv",
+  origin = "https://git.io/vb3Xa"
+)
+```
+
+Read and display some data
+
+```{r}
+pep_dat <- read_tsv(file = pep_file)
+pep_dat %>% head(5)
+```
+
+Summarize
+
+```{r}
+pep_dat %>%
+  group_by(label_chr, data_type) %>%
+  tally() %>% 
+  arrange(desc(n))
+```
+
+Plot sequence logo
+
+```{r}
+pep_dat %>%
+  filter(label_chr == "SB") %>%
+  pull(peptide) %>%
+  ggseqlogo()
+```
+
+Plot peptide image
+
+```{r}
+pep_dat %>%
+  filter(label_chr == "SB") %>%
+  head(1) %>%
+  pull(peptide) %>%
+  pep_plot_images()
+```

immunotherapy.Rproj

@@ -0,0 +1,16 @@
+Version: 1.0
+
+RestoreWorkspace: Default
+SaveWorkspace: Default
+AlwaysSaveHistory: Default
+
+EnableCodeIndexing: Yes
+UseSpacesForTab: Yes
+NumSpacesForTab: 2
+Encoding: UTF-8
+
+RnwWeave: Sweave
+LaTeX: pdfLaTeX
+
+AutoAppendNewline: Yes
+StripTrailingWhitespace: Yes

immunotherapy_notebook.Rmd

@@ -0,0 +1,71 @@
+---
+title: "Immunotherapy Notebook"
+output:
+  html_document:
+    df_print: paged
+---
+
+## Introduction
+
+Inspired by the blog post at https://blogs.rstudio.com/tensorflow/posts/2018-01-29-dl-for-cancer-immunotherapy/
+
+## Packages
+
+Packages for sequence logos and peptides
+
+```r
+devtools::install_github("omarwagih/ggseqlogo")
+devtools::install_github("leonjessen/PepTools")
+```
+
+## Analysis
+
+```{r}
+library(keras)
+library(tidyverse)
+library(PepTools)
+```
+
+Download and cache the data
+
+```{r}
+pep_file <- get_file(
+  "ran_peps_netMHCpan40_predicted_A0201_reduced_cleaned_balanced.tsv",
+  origin = "https://git.io/vb3Xa"
+)
+```
+
+Read and display some data
+
+```{r}
+pep_dat <- read_tsv(file = pep_file)
+pep_dat %>% head(5)
+```
+
+Summarize
+
+```{r}
+pep_dat %>%
+  group_by(label_chr, data_type) %>%
+  tally() %>% 
+  arrange(desc(n))
+```
+
+Plot sequence logo
+
+```{r}
+pep_dat %>%
+  filter(label_chr == "SB") %>%
+  pull(peptide) %>%
+  ggseqlogo()
+```
+
+Plot peptide image
+
+```{r}
+pep_dat %>%
+  filter(label_chr == "SB") %>%
+  head(1) %>%
+  pull(peptide) %>%
+  pep_plot_images()
+```

train_model.R

@@ -0,0 +1,129 @@
+# https://blogs.rstudio.com/tensorflow/posts/2018-01-29-dl-for-cancer-immunotherapy/
+
+# # Keras + TensorFlow and it's dependencies
+# install.packages("keras")
+# library(keras)
+# install_keras()
+#
+# # Tidyverse (readr, ggplot2, etc.)
+# install.packages("tidyverse")
+#
+
+library(keras)
+library(tfdeploy)
+library(tidyverse)
+library(PepTools)
+
+pep_file <- get_file(
+  "ran_peps_netMHCpan40_predicted_A0201_reduced_cleaned_balanced.tsv",
+  origin = "https://git.io/vb3Xa"
+)
+pep_dat <- read_tsv(file = pep_file)
+
+pep_dat %>% head()
+
+
+x_train <- pep_dat %>%
+  filter(data_type == "train") %>%
+  pull(peptide) %>%
+  pep_encode()
+
+y_train <- pep_dat %>%
+  filter(data_type == "train") %>%
+  pull(label_num) %>%
+  array()
+
+x_test <- pep_dat %>%
+  filter(data_type == "test") %>%
+  pull(peptide) %>%
+  pep_encode()
+
+y_test <- pep_dat %>%
+  filter(data_type == "test") %>%
+  pull(label_num) %>%
+  array()
+
+dim(x_train)
+
+
+x_train <- array_reshape(x_train, c(nrow(x_train), 9 * 20))
+x_test <- array_reshape(x_test, c(nrow(x_test), 9 * 20))
+
+y_train <- to_categorical(y_train, num_classes = 3)
+y_test <- to_categorical(y_test, num_classes = 3)
+
+model <- keras_model_sequential() %>%
+  layer_dense(units = 180, activation = "relu", input_shape = 180) %>%
+  layer_dropout(rate = 0.4) %>%
+  layer_dense(units = 90, activation = "relu") %>%
+  layer_dropout(rate = 0.3) %>%
+  layer_dense(units = 3, activation = "softmax")
+
+summary(model)
+
+model %>%
+  compile(
+    loss = "categorical_crossentropy",
+    optimizer = optimizer_rmsprop(),
+    metrics = c("accuracy")
+  )
+
+history <- model %>%
+  fit(
+    x_train, y_train,
+    epochs = 20,
+    batch_size = 64,
+    validation_split = 0.2
+  )
+
+
+
+library(ggplot2)
+plot(history)
+
+perf <- model %>%
+  evaluate(x_test, y_test)
+perf
+
+
+y_pred <- model %>%
+  predict_classes(x_test)
+
+y_real <- y_test %>%
+  apply(1, function(x) {
+    return(which(x == 1) - 1)
+  })
+
+results <- tibble(
+  y_real = y_real %>% factor(levels = 2:0),
+  y_pred = y_pred %>% factor(levels = 0:2),
+  Correct = if_else(y_real == y_pred, "yes", "no") %>% factor()
+)
+
+library(glue)
+
+results %>%
+  ggplot(aes(x = y_pred, y = y_real, fill = Correct)) +
+  # geom_point() +
+  geom_tile(aes(fill = Correct), stat = "sum") +
+  geom_jitter(alpha = 0.5) +
+  ggtitle(
+    label = "Performance on 10% unseen data - Feed Forward Neural Network",
+    subtitle = glue::glue("Accuracy = {round(perf$acc, 3) * 100}%")
+  ) +
+  xlab(
+    "Measured\n(Real class, as predicted by netMHCpan-4.0)"
+  ) +
+  ylab(
+    "Predicted\n(Class assigned by Keras / TensorFlow model)"
+  ) +
+  scale_fill_manual(
+    labels = c("No", "Yes"),
+    values = c("red", "blue")
+  ) +
+  theme_bw()
+
+
+# save model for deployment -----------------------------------------------
+
+model %>% tfdeploy::export_savedmodel("saved_models")