def validate_metadata(metadata_path, environmental_metadata_path): # Load metadata via QIIME2 metadata API # It will verify metadata vadlity as well try: metadata_df = load_metadata(metadata_path) except MetadataFileError as err: message = str(err) return 400, message try: environmental_metadata_df = load_metadata(environmental_metadata_path) except MetadataFileError as err: message = str(err) return 400, message # Check user-specified columns actually exist in the metadata file try: check_column_exists(metadata_df, fill_variable) except AXIOME3PipelineError as err: message = str(err) return 400, message return 200, "Ok"
def prep_bubbleplot(feature_table_artifact_path,
taxonomy_artifact_path,
metadata_path=None,
level="asv",
groupby_taxa="phylum",
abundance_threshold=0.1,
keyword=None):
feature_table_artifact = check_artifact_type(feature_table_artifact_path,
"feature_table")
taxonomy_artifact = check_artifact_type(taxonomy_artifact_path, "taxonomy")
collapsed_df = collapse_taxa(feature_table_artifact, taxonomy_artifact,
level)
original_taxa = pd.Series(collapsed_df["Taxon"])
row_id = pd.Series(collapsed_df.index)
renamed_taxa = rename_taxa(original_taxa, row_id)
percent_df = calculate_percent_value(
collapsed_df.drop(["Taxon"], axis="columns"))
percent_df["SpeciesName"] = renamed_taxa
taxa_group = group_by_taxa(original_taxa, groupby_taxa, level)
percent_df["TaxaGroup"] = taxa_group
filter_criteria = filter_by_keyword(original_taxa, keyword)
filtered_df = percent_df.loc[filter_criteria, ]
long_df = pd.melt(filtered_df,
id_vars=['SpeciesName', 'TaxaGroup'],
var_name="SampleName",
value_name="Percentage")
abundance_filtered_df = filter_by_abundance(long_df, "Percentage",
abundance_threshold)
rounded_abundance_filtered_df = round_percentage(abundance_filtered_df,
"Percentage", 3)
sorted_df = alphabetical_sort_df(rounded_abundance_filtered_df,
["TaxaGroup", "SpeciesName"])
# Make SpeciesName column category to avoid automatic sorting
sorted_df['SpeciesName'] = pd.Categorical(
sorted_df['SpeciesName'],
categories=sorted_df['SpeciesName'].unique(),
ordered=True)
# Join metadata with bubbleplot df
if (metadata_path is not None):
metadata_df = load_metadata(metadata_path)
merged_df = sorted_df.merge(metadata_df,
how="inner",
left_on="SampleName",
right_index=True)
return merged_df
return sorted_df
def validate_metadata(metadata_path, target_primary, target_secondary): # Load metadata via QIIME2 metadata API # It will verify metadata vadlity as well try: metadata_df = load_metadata(metadata_path) except MetadataFileError as err: message = str(err) return 400, message # Check user-specified columns actually exist in the metadata file try: check_column_exists(metadata_df, target_primary, target_secondary) except AXIOME3PipelineError as err: message = str(err) return 400, message return 200, "Ok"
def generate_pcoa_plot(pcoa,
metadata,
colouring_variable,
shape_variable=None,
primary_dtype="category",
secondary_dtype="category",
palette='Paired',
brewer_type='qual',
alpha=0.9,
stroke=0.6,
point_size=6,
x_axis_text_size=10,
y_axis_text_size=10,
legend_title_size=10,
legend_text_size=10,
PC_axis1=1,
PC_axis2=2):
# raise AXIOME3Error if PC_axis1 == PC_axis2
if (PC_axis1 == PC_axis2):
raise AXIOME3Error("PC axis one and PC axis two cannot be equal!")
# Load metadata file
metadata_df = load_metadata(metadata)
# Inner join metadata file with ordinations
pcoa_coords = pcoa.samples
pcoa_data_samples = pd.merge(pcoa_coords,
right=metadata_df,
left_index=True,
right_index=True)
# Make x and y axis labels
proportions = pcoa.proportion_explained
x_explained_idx = PC_axis1 - 1
y_explained_idx = PC_axis2 - 1
pc_1 = 'Axis ' + str(PC_axis1)
pc_2 = 'Axis ' + str(PC_axis2)
x_explained = str(round(proportions[x_explained_idx] * 100, 1))
y_explained = str(round(proportions[y_explained_idx] * 100, 1))
# Convert user specified columns to category
# **BIG ASSUMPTION HERE**
pcoa_data_samples = convert_col_dtype(pcoa_data_samples,
colouring_variable, primary_dtype)
if (shape_variable is not None):
pcoa_data_samples = convert_col_dtype(pcoa_data_samples,
shape_variable, secondary_dtype)
# Pre-format target variables
#primary_target_fill = 'factor(' + str(colouring_variable) + ')'
primary_target_fill = str(colouring_variable)
if (shape_variable is not None):
secondary_target_fill = str(shape_variable)
ggplot_obj = ggplot(
pcoa_data_samples,
aes(x=pc_1,
y=pc_2,
fill=primary_target_fill,
shape=secondary_target_fill))
else:
ggplot_obj = ggplot(pcoa_data_samples,
aes(x=pc_1, y=pc_2, fill=primary_target_fill))
# Plot the data
pcoa_plot = (
ggplot_obj + geom_point(size=point_size, alpha=alpha, stroke=stroke) +
theme_bw() +
theme(panel_grid=element_blank(),
line=element_line(colour='black'),
panel_border=element_rect(colour='black'),
legend_title=element_text(size=legend_title_size, face='bold'),
legend_key=element_blank(),
legend_text=element_text(size=legend_text_size),
axis_title_x=element_text(size=x_axis_text_size),
axis_title_y=element_text(size=y_axis_text_size),
legend_key_height=5,
text=element_text(family='Arial', colour='black')) +
xlab(pc_1 + ' (' + x_explained + '%)') +
ylab(pc_2 + ' (' + y_explained + '%)'))
# Custom colours
color_len = len(pcoa_data_samples[colouring_variable].unique())
color_name = str(colouring_variable)
pcoa_plot = add_fill_colours_from_users(pcoa_plot, color_name, palette,
brewer_type)
# Custom shapes
if (shape_variable is not None):
shape_len = len(pcoa_data_samples[shape_variable].unique())
shape_name = str(shape_variable)
pcoa_plot = add_discrete_shape(pcoa_plot, shape_len, shape_name)
return pcoa_plot
if __name__ == "__main__":
parser = args_parse()
# Print help messages if no arguments are supplied
if (len(sys.argv) < 2):
parser.print_help()
sys.exit(0)
args = parser.parse_args()
# Load QIIME2 PCoA result and convert to PCoA object
pcoa = convert_qiime2_2_skbio(args.pcoa_qza)
# Load metadata
metadata_df = load_metadata(args.metadata)
# Check if metadata has target columns
check_column_exists(metadata_df, args.target_primary,
args.target_secondary)
# Generate PCoA plot
pcoa_plot = generate_pcoa_plot(pcoa=pcoa,
metadata=args.metadata,
colouring_variable=args.target_primary,
shape_variable=args.target_secondary,
point_size=args.point_size,
alpha=args.alpha,
stroke=args.stroke,
PC_axis1=args.pc_axis_one,
PC_axis2=args.pc_axis_two)
def prep_triplot_input(sample_metadata_path,
env_metadata_path,
feature_table_artifact_path,
taxonomy_artifact_path,
sampling_depth=0,
ordination_collapse_level="asv",
wascores_collapse_level="phylum",
dissmilarity_index="Bray-Curtis",
R2_threshold=0.1,
pval_threshold=0.05,
wa_threshold=0.1,
PC_axis_one=1,
PC_axis_two=2,
output_dir='.'):
# Load sample metadata
sample_metadata_df = load_metadata(sample_metadata_path)
# Load environmental metadata
# and drop rows with missing values (WARN users?)
env_metadata_df = load_env_metadata(env_metadata_path)
env_metadata_df = env_metadata_df.dropna()
# Load feature table and collapse
feature_table_artifact = check_artifact_type(feature_table_artifact_path,
"feature_table")
taxonomy_artifact = check_artifact_type(taxonomy_artifact_path, "taxonomy")
ordination_collapsed_df = collapse_taxa(feature_table_artifact,
taxonomy_artifact, sampling_depth,
ordination_collapse_level)
abundance_collapsed_df = collapse_taxa(feature_table_artifact,
taxonomy_artifact, sampling_depth,
wascores_collapse_level)
# Rename taxa for wascores collapsed df
original_taxa = pd.Series(abundance_collapsed_df["Taxon"])
row_id = pd.Series(abundance_collapsed_df.index)
renamed_taxa = rename_taxa(original_taxa, row_id)
abundance_collapsed_df['Taxa'] = renamed_taxa
abundance_collapsed_df_reindexed = abundance_collapsed_df.set_index('Taxa')
abundance_collapsed_df_reindexed = abundance_collapsed_df_reindexed.drop(
['Taxon'], axis="columns")
# transpose feature table so that it has samples as rows, taxa/ASV as columns
ordination_transposed_df = ordination_collapsed_df.drop(['Taxon'],
axis="columns").T
abundance_transposed_df = abundance_collapsed_df_reindexed.T
# Remove samples that have total counts <= 5 (R complains if total count <= 5)
count_filtered_df = filter_by_total_count(ordination_transposed_df)
# Find sample intersection of feature table, sample metadata, and environmental metadata
intersection_feature_table_df, intersection_abundance_df, intersection_sample_metadata_df, intersection_environmental_metadata_df, sample_summary = find_sample_intersection(
count_filtered_df, abundance_transposed_df, sample_metadata_df,
env_metadata_df)
process_input_with_R(intersection_feature_table_df,
intersection_abundance_df,
intersection_sample_metadata_df,
intersection_environmental_metadata_df,
VEGDIST_OPTIONS[dissmilarity_index], R2_threshold,
pval_threshold, wa_threshold, PC_axis_one,
PC_axis_two, output_dir)
# After successful subprocess call, there should be intermediate output files...
# Elegant way to do this?
merged_df_path = os.path.join(output_dir, "processed_merged_df.csv")
proj_arrow_df_path = os.path.join(output_dir,
"processed_vector_arrow_df.csv")
filtered_wa_df_path = os.path.join(output_dir, "processed_wa_df.csv")
proj_env_df_path = os.path.join(output_dir, "processed_projection_df.csv")
proportion_df_path = os.path.join(output_dir,
"processed_proportion_explained_df.csv")
# Read back into pandas dataframes
merged_df = pd.read_csv(merged_df_path)
renamed_vector_arrow_df = pd.read_csv(proj_arrow_df_path)
filtered_wascores_df = pd.read_csv(filtered_wa_df_path)
proportion_explained = pd.read_csv(proportion_df_path)
projection_df = pd.read_csv(proj_env_df_path)
return merged_df, renamed_vector_arrow_df, filtered_wascores_df, proportion_explained, projection_df, sample_summary