Skip to content

Commit

Permalink
Version 11 - Complete
Browse files Browse the repository at this point in the history 
V11 (Mar 14, 2024): Some more exceptions handled. When the clinical and RNA patient IDs do not match, and when there are NaNs in the time to event column. The logging has not been revised yet.
  • Loading branch information
@EdRey05
EdRey05 authored Mar 14, 2024
1 parent 5f31d4b commit 9679d19
Showing 1 changed file with 58 additions and 32 deletions.
90 changes: 58 additions & 32 deletions 003_KM_plotter/app.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -3,11 +3,11 @@
Eduardo Reyes Alvarez, Ph.D.
Contact:
[email protected]
Done: CS, FEZF1, GRK7, MSH5, NOS3, NTHL1, RUVBL1, SELPLG, SEMA3A, TNK2, UBR5
App version:
V10 (Jan 06, 2024): Major improvements to -Using variable(s)- option. Now the content of each
selected column/variable is analyzed better in base of 9 observed scenarios
to display the most appropriate plot and subgrouping widgets. The logging
V11 (Mar 14, 2024): Some more exceptions handled. When the clinical and RNA patient IDs do not
match, and when there are NaNs in the time to event column. The logging
has not been revised yet.
'''
###################################################################################################
Expand Down Expand Up @@ -42,16 +42,16 @@

# Use these colors for altair charts
st.session_state["alt_colors"] = ["#76448A", "#B03A2E", "#1E8449", "#1F618D", "#34495E ",
"#D68910", "#707B7C", "#E67E22", "#2E86C1", "#E74C3C",
"#2C3E50", "#F1C40F", "#3498DB", "#D35400", "#27AE60"]
"#D68910", "#707B7C", "#E67E22", "#2E86C1", "#E74C3C",
"#2C3E50", "#F1C40F", "#3498DB", "#D35400", "#27AE60"]
# Title
st.title("Interactive Kaplan-Meier plot generator")
st.markdown('<hr style="margin-top: +2px; margin-bottom: +2px; border-width: 5px;">', unsafe_allow_html=True)

# Sidebar - Initial widgets
with st.sidebar:
uploaded_files = st.file_uploader(label="Upload a clinical file (and optionally, a RNA file)",
type=["txt"], accept_multiple_files=True)
type=["txt"], accept_multiple_files=True)
st.markdown('<hr style="margin-top: 1px; margin-bottom: 1px; border-width: 5px;">', unsafe_allow_html=True)
start_button = st.button(label="Begin", type="secondary")
restart_button = st.button(label="Start over", type="secondary")
Expand All @@ -60,9 +60,9 @@

# Function to setup the logging configuration
def logging_setup():
# Check if logging has already been initialized
if "log_created" not in st.session_state:

# Check if logging has already been initialized
if "log_created" not in st.session_state:

# Configure the logging settings
logging.basicConfig(level=logging.INFO, format="%(asctime)s - %(levelname)s - %(message)s")
Expand Down Expand Up @@ -99,7 +99,7 @@ def load_input_files(uploaded_files):
# Check if one or two files were uploaded
clinical_file = next((file for file in uploaded_files if file.name == "clinical.txt"), None)
RNA_file = next((file for file in uploaded_files if file.name == "RNA.txt"), None)

# Load the files available
if clinical_file is not None:

Expand Down Expand Up @@ -208,9 +208,9 @@ def file_preprocessing():

# Save the neccesary data for the next steps in the session state
st.session_state.update({"df_clinical": df_clinical,
"df_RNA": df_RNA,
"time_to_event_options": time_to_event_options,
"event_observation_options": event_observation_options})
"df_RNA": df_RNA,
"time_to_event_options": time_to_event_options,
"event_observation_options": event_observation_options})

# Also add the options of genes if a RNA file was uploaded
if df_RNA is not None:
Expand Down Expand Up @@ -239,10 +239,10 @@ def widget_preparation():

# Save the columns and containers in the session state
widget_and_output_areas = [col_1_row_1, col_2_row_1,
col_1_row_2, col_2_row_2,
col_1_row_3, col_2_row_3, col_3_row_3, col_4_row_3,
col_1_row_14, col_2_row_14, col_3_row_14,
col_1_row_15, col_2_row_15, col_3_row_15, col_4_row_15]
col_1_row_2, col_2_row_2,
col_1_row_3, col_2_row_3, col_3_row_3, col_4_row_3,
col_1_row_14, col_2_row_14, col_3_row_14,
col_1_row_15, col_2_row_15, col_3_row_15, col_4_row_15]
st.session_state["widget_and_output_areas"] = widget_and_output_areas

# Time to event widget and callback function
Expand Down Expand Up @@ -341,26 +341,31 @@ def time_to_event_dropdown_handler(change):
time_column = df_clinical[column_name].dropna()
logger.info(f"The user selected: {column_name} Widget: time_to_event_dropdown. \n")
logger.info(f"Original dtype of {column_name}: {df_clinical[column_name].dtype} Dtype once removing NANs: {time_column.dtype} \n")

# Make histogram of values with altair and handle exceptions
if time_column.dtype != "object":
alt_data1 = pd.DataFrame({column_name: time_column})
if time_column.dtype == "object":
# First try to convert the column to a numeric type and if it fails we show a warning
try:
time_column = pd.to_numeric(time_column, errors="coerce")
time_column = time_column.dropna()
except:
info_str = "Warning: Column type is not numeric."
logger.warning("User attention required: The time to event column may not be numerical. \n")

with time_to_event_output:
time_to_event_output.empty()
st.warning(info_str)
st.stop()

chart1 = alt.Chart(alt_data1).mark_bar(color="#BA4A00").encode(
alt_data1 = pd.DataFrame({column_name: time_column})

chart1 = alt.Chart(alt_data1).mark_bar(color="#BA4A00").encode(
alt.X(column_name, type="quantitative", bin=alt.Bin(step=12)),
alt.Y("count()", title="Patients"),
).properties(width=425, height=325
).configure_axis(labelColor="#3386BD")
logger.info(f"A histogram was successfully made and displayed for: {column_name} \n")
else:
# If the column is not numeric, display a warning and stop the app
info_str = "Warning: Column type is not numeric."
logger.warning("User attention required: The time to event column may not be numerical. \n")
logger.info(f"A histogram was successfully made and displayed for: {column_name} \n")

with time_to_event_output:
time_to_event_output.empty()
st.warning(info_str)
st.stop()
else:
# If the column is not in the df, display an error and stop the app
info_str = "Warning: Column not found in the dataframe."
Expand Down Expand Up @@ -496,7 +501,7 @@ def subgroup_buttons_handler(change):
# Show a slider for 1 to 5 variables (to prevent crazy number of curves)
with subgroup_buttons_output:
variable_number_slider = st.slider(label="Number of variables:",
min_value=1, max_value=5, step=1, value=1)
min_value=1, max_value=5, step=1, value=1)
variable_number_slider_handler(variable_number_slider)
else:
# If the user doesn't want to make subgroups, don't show the slider
Expand Down Expand Up @@ -1317,6 +1322,13 @@ def variables_selection_handler(change, repeat):
logger.info(f"[Subgrouping 2nd step] The column {change} -{KM_data_1var.dtypes[change]} dtype- from df_RNA was selected to make subgroups. \n")
column_data[repeat - 1] = KM_data_1var[change].copy()

# Raise a warning when the patient IDs do not match in the clinical and RNA datasets, causing an empty df
if len(KM_data_1var) == 0:
logger.warning(f"[Subgrouping 2nd step] The merge of datasets produced an empty dataframe, it is likely that the IDs may not match. \n")
st.error("The merge of datasets produced an empty dataframe, check that the patient IDs match in both datasets. \n")
st.error("You may need to check the data source or pre-process the datasets to make the IDs consistent before using this app.")
st.stop()

# Log the current status of KM_data_1var
logger.info(f"[Subgrouping 2nd step] Keep relevant columns of KM_data_1var and only rows with 0/1 event labels: \n {KM_data_1var.iloc[:15, :10].to_string()} \n")
logger.info(f"[Subgrouping 2nd step] Data types of KM_data_1var columns: \n {KM_data_1var.dtypes.to_string()} \n\n")
Expand Down Expand Up @@ -1454,6 +1466,16 @@ def pass_KM_parameters():
KM_plot_area.empty()
st.warning("First select the values to label as 0 and 1 (No event, event)!!")
st.stop()

# Check if the time to event column is numerical
if df_clinical[time_to_event_selection].dtype == "object":
# First try to convert the column to a numeric type and if it fails we show a warning
try:
df_clinical[time_to_event_selection] = pd.to_numeric(df_clinical[time_to_event_selection], errors="coerce")
df_clinical[time_to_event_selection] = df_clinical[time_to_event_selection].dropna()
except:
st.warning("Warning: Time to Event column is not numeric.")
st.stop()

# If no subgrouping is required, apply the event tags and pass the data to KM_analysis
if subgroup_buttons_selection == "None":
Expand Down Expand Up @@ -1700,6 +1722,10 @@ def KM_analysis(KM_data, KM_subgroups):

# Create KaplanMeierFitter objects for each subgroup in KM_subgroups
for label, subset in KM_subgroups.items():

# Remove any rows in the subset with NaN values
subset = subset.dropna()

kmf = KaplanMeierFitter()
kmf.fit(durations=subset[current_time_column], event_observed=subset[current_event_column])
KMF_object[label] = kmf
Expand Down Expand Up @@ -1741,7 +1767,7 @@ def save_KM_results(generate_plot_button):
# Prepare the data to be processed (single KM object or list of KM objects)
if isinstance(KM_analysis_output, dict):
KM_objects_to_process = [{"label": f"KM_Subgroup_{i+1}", "KM_object": KM_object}
for i, (label, KM_object) in enumerate(KM_analysis_output.items())]
for i, (label, KM_object) in enumerate(KM_analysis_output.items())]
real_labels = [f"KM_Subgroup_{i+1}: {label}" for i, (label, KM_object) in enumerate(KM_analysis_output.items())]
else:
# If KM_analysis_output is a single KM object, add it to the list as a dictionary with a general label
Expand Down

0 comments on commit 9679d19

Please sign in to comment.