masster 0.4.22__py3-none-any.whl → 0.5.0__py3-none-any.whl

masster/study/parameters.py

@@ -8,7 +8,7 @@ similar to the sample parameters module but for study-level operations.
 from __future__ import annotations
 
 
-def store_history(self, keys, value):
+def update_history(self, keys, value):
     """
     Store parameters in a nested dictionary structure.
 
@@ -74,10 +74,10 @@ def update_parameters(self, **kwargs):
         # Check if it's a parameter defaults instance
         if hasattr(value, "to_dict") and callable(getattr(value, "to_dict")):
             # Store the parameter object
-            self.store_history([key], value.to_dict())
+            self.update_history([key], value.to_dict())
         else:
             # Store individual parameter
-            self.store_history([key], value)
+            self.update_history([key], value)
 
 
 def get_parameters_property(self):

masster/study/plot.py

@@ -308,7 +308,7 @@ def plot_alignment(
         self.logger.info("Showing current RT values for both plots. Run align() first to see alignment comparison.")
 
     # Get sample_uids to filter by if specified
-    sample_uids = self._get_sample_uids(samples) if samples is not None else None
+    sample_uids = self._get_samples_uids(samples) if samples is not None else None
 
     # Start with full features_df
     features_df = self.features_df
@@ -836,7 +836,7 @@ def plot_samples_2d(
     from bokeh.io.export import export_png
     from bokeh.models import ColumnDataSource, HoverTool
 
-    sample_uids = self._get_sample_uids(samples)
+    sample_uids = self._get_samples_uids(samples)
 
     if not sample_uids:
         self.logger.error("No valid sample_uids provided.")
@@ -1053,7 +1053,7 @@ def plot_bpc(
     from bokeh.io.export import export_png
     from masster.study.helpers import get_bpc
 
-    sample_uids = self._get_sample_uids(samples)
+    sample_uids = self._get_samples_uids(samples)
     if not sample_uids:
         self.logger.error("No valid sample_uids provided for BPC plotting.")
         return
@@ -1238,7 +1238,7 @@ def plot_eic(
         self.logger.error("mz must be provided for EIC plotting")
         return
 
-    sample_uids = self._get_sample_uids(samples)
+    sample_uids = self._get_samples_uids(samples)
     if not sample_uids:
         self.logger.error("No valid sample_uids provided for EIC plotting.")
         return
@@ -1400,7 +1400,7 @@ def plot_rt_correction(
         self.logger.error("Column 'rt_original' not found in features_df. Alignment/backup RTs missing.")
         return
 
-    sample_uids = self._get_sample_uids(samples)
+    sample_uids = self._get_samples_uids(samples)
     if not sample_uids:
         self.logger.error("No valid sample_uids provided for RT correction plotting.")
         return
@@ -1537,7 +1537,7 @@ def plot_chrom(
     height=300,
 ):
     cons_uids = self._get_consensus_uids(uids)
-    sample_uids = self._get_sample_uids(samples)
+    sample_uids = self._get_samples_uids(samples)
 
     chroms = self.get_chrom(uids=cons_uids, samples=sample_uids)
 
@@ -1723,18 +1723,32 @@ def plot_chrom(
 def plot_consensus_stats(
     self,
     filename=None,
-    width=1200,
+    width=840,  # Reduced from 1200 (30% smaller)
     height=None,
     alpha=0.6,
     bins=30,
     n_cols=4,
 ):
     """
-    Plot histograms/distributions for all numeric columns in consensus_df.
+    Plot histograms/distributions for specific consensus statistics in the requested order.
+    
+    Shows the following properties in order:
+    1. rt: Retention time
+    2. rt_delta_mean: Mean retention time delta
+    3. mz: Mass-to-charge ratio
+    4. mz_range: Mass range (mz_max - mz_min)
+    5. log10_inty_mean: Log10 of mean intensity
+    6. number_samples: Number of samples
+    7. number_ms2: Number of MS2 spectra
+    8. charge_mean: Mean charge
+    9. quality: Feature quality
+    10. chrom_coherence_mean: Mean chromatographic coherence
+    11. chrom_height_scaled_mean: Mean scaled chromatographic height
+    12. chrom_prominence_scaled_mean: Mean scaled chromatographic prominence
     
     Parameters:
         filename (str, optional): Output filename for saving the plot
-        width (int): Overall width of the plot (default: 1200)
+        width (int): Overall width of the plot (default: 840)
         height (int, optional): Overall height of the plot (auto-calculated if None)
         alpha (float): Histogram transparency (default: 0.6)
         bins (int): Number of histogram bins (default: 30)
@@ -1753,24 +1767,48 @@ def plot_consensus_stats(
     # Get all columns and their data types - work with original dataframe
     data_df = self.consensus_df.clone()
 
-    # Identify numeric columns (excluding ID columns that are typically strings)
-    id_columns = ["consensus_uid", "consensus_id", "uid", "id"]
-    numeric_columns = []
+    # Define specific columns to plot in the exact order requested
+    desired_columns = [
+        "rt", 
+        "rt_delta_mean", 
+        "mz", 
+        "mz_range",  # mz_max-mz_min (will be calculated)
+        "log10_inty_mean",  # log10(inty_mean) (will be calculated)
+        "number_samples", 
+        "number_ms2", 
+        "charge_mean", 
+        "quality", 
+        "chrom_coherence_mean", 
+        "chrom_height_scaled_mean", 
+        "chrom_prominence_scaled_mean"
+    ]
+    
+    # Calculate derived columns if they don't exist
+    if "mz_range" not in data_df.columns and "mz_max" in data_df.columns and "mz_min" in data_df.columns:
+        data_df = data_df.with_columns((pl.col("mz_max") - pl.col("mz_min")).alias("mz_range"))
+    
+    if "log10_inty_mean" not in data_df.columns and "inty_mean" in data_df.columns:
+        data_df = data_df.with_columns(pl.col("inty_mean").log10().alias("log10_inty_mean"))
     
-    for col in data_df.columns:
-        if col not in id_columns:
-            dtype = data_df[col].dtype
-            # Check if column is numeric (int, float, or can be converted to numeric)
-            if dtype in [pl.Int8, pl.Int16, pl.Int32, pl.Int64, 
-                        pl.UInt8, pl.UInt16, pl.UInt32, pl.UInt64,
-                        pl.Float32, pl.Float64]:
-                numeric_columns.append(col)
+    # Filter to only include columns that exist in the dataframe, preserving order
+    numeric_columns = [col for col in desired_columns if col in data_df.columns]
+    
+    # Check if the numeric columns are actually numeric
+    final_numeric_columns = []
+    for col in numeric_columns:
+        dtype = data_df[col].dtype
+        if dtype in [pl.Int8, pl.Int16, pl.Int32, pl.Int64, 
+                    pl.UInt8, pl.UInt16, pl.UInt32, pl.UInt64,
+                    pl.Float32, pl.Float64]:
+            final_numeric_columns.append(col)
+    
+    numeric_columns = final_numeric_columns
 
     if len(numeric_columns) == 0:
-        self.logger.error("No numeric columns found in consensus_df for plotting distributions.")
+        self.logger.error(f"None of the requested consensus statistics columns were found or are numeric. Available columns: {list(data_df.columns)}")
         return
 
-    self.logger.debug(f"Creating distribution plots for {len(numeric_columns)} numeric columns: {numeric_columns}")
+    self.logger.debug(f"Creating distribution plots for {len(numeric_columns)} specific consensus columns: {numeric_columns}")
 
     # Work directly with Polars - no conversion to pandas needed
     data_df_clean = data_df.select(numeric_columns)
@@ -1798,25 +1836,29 @@ def plot_consensus_stats(
     
     # Auto-calculate height if not provided
     if height is None:
-        plot_height = 300
-        height = plot_height * n_rows + 100  # Add some padding
+        plot_height = 210  # Reduced from 300 (30% smaller)
+        height = plot_height * n_rows + 56  # Reduced from 80 (30% smaller)
     else:
-        plot_height = (height - 100) // n_rows  # Subtract padding and divide
+        plot_height = (height - 56) // n_rows  # Reduced padding (30% smaller)
     
-    plot_width = (width - 100) // n_cols  # Subtract padding and divide
+    plot_width = (width - 56) // n_cols  # Reduced padding (30% smaller)
 
     # Create plots grid
     plots = []
     current_row = []
     
     for i, col in enumerate(numeric_columns):
+        # Check if this column should use log scale for y-axis
+        y_axis_type = "log" if col in ["number_samples", "number_ms2"] else "linear"
+        
         # Create histogram for this column
         p = figure(
             width=plot_width,
             height=plot_height,
             title=col,
             toolbar_location="above",
-            tools="pan,wheel_zoom,box_zoom,reset,save"
+            tools="pan,wheel_zoom,box_zoom,reset,save",
+            y_axis_type=y_axis_type
         )
         
         # Set white background
@@ -1840,10 +1882,19 @@ def plot_consensus_stats(
         values_array = valid_values.to_numpy()
         hist, edges = np.histogram(values_array, bins=bins)
         
+        # Handle log y-axis: replace zero counts with small positive values
+        if y_axis_type == "log":
+            # Replace zero counts with a small value (1e-1) to make them visible on log scale
+            hist_log_safe = np.where(hist == 0, 0.1, hist)
+            bottom_val = 0.1  # Use small positive value for bottom on log scale
+        else:
+            hist_log_safe = hist
+            bottom_val = 0
+        
         # Create histogram bars
         p.quad(
-            top=hist,
-            bottom=0,
+            top=hist_log_safe,
+            bottom=bottom_val,
             left=edges[:-1],
             right=edges[1:],
             fill_color="steelblue",
@@ -1852,11 +1903,16 @@ def plot_consensus_stats(
         )
         
         # Style the plot
-        p.title.text_font_size = "12pt"
-        p.xaxis.axis_label = col
-        p.yaxis.axis_label = "Count"
-        p.grid.visible = True
-        p.grid.grid_line_color = "#E0E0E0"
+        p.title.text_font_size = "10pt"  # Reduced from 12pt
+        p.xaxis.axis_label = ""  # Remove x-axis title
+        p.grid.grid_line_alpha = 0.3  # Show y-axis grid with transparency
+        p.grid.grid_line_color = "gray"
+        p.grid.grid_line_dash = [6, 4]  # Dashed grid lines
+        p.xgrid.visible = False  # Hide x-axis grid
+        p.outline_line_color = None  # Remove gray border around plot area
+        
+        # Remove y-axis label but keep y-axis visible
+        p.yaxis.axis_label = ""
         
         current_row.append(p)
         
@@ -1868,14 +1924,12 @@ def plot_consensus_stats(
             plots.append(current_row)
             current_row = []
 
-    # Create grid layout
-    grid = gridplot(plots)
+    # Create grid layout with white background
+    grid = gridplot(plots, toolbar_location="above", merge_tools=True)
     
-    # Set overall background to white
-    if hasattr(grid, "background_fill_color"):
-        grid.background_fill_color = "white"
-    if hasattr(grid, "border_fill_color"):
-        grid.border_fill_color = "white"
+    # The background should be white by default in Bokeh
+    # Individual plots already have white backgrounds set above
+
 
     # Apply consistent save/display behavior
     if filename is not None:
@@ -1895,7 +1949,7 @@ def plot_consensus_stats(
     return grid
 
 
-def plot_pca(
+def plot_samples_pca(
     self,
     filename=None,
     width=500,
@@ -2035,6 +2089,7 @@ def plot_pca(
         tools="pan,wheel_zoom,box_zoom,reset,save",
     )
 
+    p.grid.visible = False
     p.xaxis.axis_label = f"PC1 ({explained_var[0]:.1%} variance)"
     p.yaxis.axis_label = f"PC2 ({explained_var[1]:.1%} variance)"
 
@@ -2159,6 +2214,293 @@ def plot_pca(
     return p
 
 
+def plot_samples_umap(
+    self,
+    filename=None,
+    width=500,
+    height=450,
+    alpha=0.8,
+    markersize=6,
+    n_components=2,
+    colorby=None,
+    title="UMAP of Consensus Matrix",
+    n_neighbors=15,
+    min_dist=0.1,
+    metric="euclidean",
+    random_state=42,
+):
+    """
+    Plot UMAP (Uniform Manifold Approximation and Projection) of the consensus matrix using Bokeh.
+
+    Parameters:
+        filename (str, optional): Output filename for saving the plot
+        width (int): Plot width (default: 500)
+        height (int): Plot height (default: 450)
+        alpha (float): Point transparency (default: 0.8)
+        markersize (int): Size of points (default: 6)
+        n_components (int): Number of UMAP components to compute (default: 2)
+        colorby (str, optional): Column from samples_df to color points by
+        title (str): Plot title (default: "UMAP of Consensus Matrix")
+        n_neighbors (int): Number of neighbors for UMAP (default: 15)
+        min_dist (float): Minimum distance for UMAP (default: 0.1)
+        metric (str): Distance metric for UMAP (default: "euclidean")
+        random_state (int or None): Random state for reproducibility (default: 42).
+            - Use an integer (e.g., 42) for reproducible results (slower, single-threaded)
+            - Use None for faster computation with multiple cores (non-reproducible)
+            
+    Note:
+        Setting random_state forces single-threaded computation but ensures reproducible results.
+        Set random_state=None to enable parallel processing for faster computation.
+    """
+    try:
+        import umap
+    except ImportError:
+        self.logger.error("UMAP not available. Please install umap-learn: pip install umap-learn")
+        return
+
+    from bokeh.models import ColumnDataSource, HoverTool, ColorBar, LinearColorMapper
+    from bokeh.plotting import figure
+    from bokeh.palettes import Category20, viridis
+    from bokeh.transform import factor_cmap
+    from sklearn.preprocessing import StandardScaler
+    import pandas as pd
+    import numpy as np
+
+    # Check if consensus matrix and samples_df exist
+    try:
+        consensus_matrix = self.get_consensus_matrix()
+        samples_df = self.samples_df
+    except Exception as e:
+        self.logger.error(f"Error getting consensus matrix or samples_df: {e}")
+        return
+
+    if consensus_matrix is None or consensus_matrix.shape[0] == 0:
+        self.logger.error("No consensus matrix available. Run merge/find_consensus first.")
+        return
+
+    if samples_df is None or samples_df.is_empty():
+        self.logger.error("No samples dataframe available.")
+        return
+
+    self.logger.debug(f"Performing UMAP on consensus matrix with shape: {consensus_matrix.shape}")
+
+    # Extract only the sample columns (exclude consensus_uid column)
+    sample_cols = [col for col in consensus_matrix.columns if col != "consensus_uid"]
+
+    # Convert consensus matrix to numpy, excluding the consensus_uid column
+    if hasattr(consensus_matrix, "select"):
+        # Polars DataFrame
+        matrix_data = consensus_matrix.select(sample_cols).to_numpy()
+    else:
+        # Pandas DataFrame or other - drop consensus_uid column
+        matrix_sample_data = consensus_matrix.drop(columns=["consensus_uid"], errors="ignore")
+        if hasattr(matrix_sample_data, "values"):
+            matrix_data = matrix_sample_data.values
+        elif hasattr(matrix_sample_data, "to_numpy"):
+            matrix_data = matrix_sample_data.to_numpy()
+        else:
+            matrix_data = np.array(matrix_sample_data)
+
+    # Transpose matrix so samples are rows and features are columns
+    matrix_data = matrix_data.T
+
+    # Handle missing values by replacing with 0
+    matrix_data = np.nan_to_num(matrix_data, nan=0.0, posinf=0.0, neginf=0.0)
+
+    # Standardize the data
+    scaler = StandardScaler()
+    matrix_scaled = scaler.fit_transform(matrix_data)
+
+    # Perform UMAP
+    reducer = umap.UMAP(
+        n_components=n_components,
+        n_neighbors=n_neighbors,
+        min_dist=min_dist,
+        metric=metric,
+        random_state=random_state,
+        n_jobs=1
+    )
+    umap_result = reducer.fit_transform(matrix_scaled)
+
+    self.logger.debug(f"UMAP completed with shape: {umap_result.shape}")
+
+    # Convert samples_df to pandas for easier manipulation
+    samples_pd = samples_df.to_pandas()
+
+    # Create dataframe with UMAP results and sample information
+    umap_df = pd.DataFrame({
+        "UMAP1": umap_result[:, 0],
+        "UMAP2": umap_result[:, 1] if n_components > 1 else np.zeros(len(umap_result)),
+    })
+
+    # Add sample information to UMAP dataframe
+    if len(samples_pd) == len(umap_df):
+        for col in samples_pd.columns:
+            umap_df[col] = samples_pd[col].values
+    else:
+        self.logger.warning(
+            f"Sample count mismatch: samples_df has {len(samples_pd)} rows, "
+            f"but consensus matrix has {len(umap_df)} samples",
+        )
+
+    # Prepare color mapping
+    color_column = None
+    color_mapper = None
+
+    if colorby and colorby in umap_df.columns:
+        color_column = colorby
+        unique_values = umap_df[colorby].unique()
+
+        # Handle categorical vs numeric coloring
+        if umap_df[colorby].dtype in ["object", "string", "category"]:
+            # Categorical coloring
+            if len(unique_values) <= 20:
+                palette = Category20[min(20, max(3, len(unique_values)))]
+            else:
+                palette = viridis(min(256, len(unique_values)))
+            color_mapper = factor_cmap(colorby, palette, unique_values)
+        else:
+            # Numeric coloring
+            palette = viridis(256)
+            color_mapper = LinearColorMapper(
+                palette=palette,
+                low=umap_df[colorby].min(),
+                high=umap_df[colorby].max(),
+            )
+
+    # Create Bokeh plot
+    p = figure(
+        width=width,
+        height=height,
+        title=f"{title}",
+        tools="pan,wheel_zoom,box_zoom,reset,save",
+    )
+
+    p.grid.visible = False
+    p.xaxis.axis_label = "UMAP1"
+    p.yaxis.axis_label = "UMAP2"
+
+    # Create data source
+    source = ColumnDataSource(umap_df)
+
+    # Create scatter plot
+    if color_mapper:
+        if isinstance(color_mapper, LinearColorMapper):
+            scatter = p.scatter(
+                "UMAP1",
+                "UMAP2",
+                size=markersize,
+                alpha=alpha,
+                color={"field": colorby, "transform": color_mapper},
+                source=source,
+            )
+            # Add colorbar for numeric coloring
+            color_bar = ColorBar(color_mapper=color_mapper, width=8, location=(0, 0))
+            p.add_layout(color_bar, "right")
+        else:
+            scatter = p.scatter(
+                "UMAP1",
+                "UMAP2",
+                size=markersize,
+                alpha=alpha,
+                color=color_mapper,
+                source=source,
+                legend_field=colorby,
+            )
+    else:
+        # If no color_by provided, use sample_color column from samples_df
+        if "sample_uid" in umap_df.columns or "sample_name" in umap_df.columns:
+            # Choose the identifier to map colors by
+            id_col = "sample_uid" if "sample_uid" in umap_df.columns else "sample_name"
+
+            # Get colors from samples_df based on the identifier
+            if id_col == "sample_uid":
+                sample_colors = (
+                    self.samples_df.filter(pl.col("sample_uid").is_in(umap_df[id_col].unique()))
+                    .select(["sample_uid", "sample_color"])
+                    .to_dict(as_series=False)
+                )
+                color_map = dict(zip(sample_colors["sample_uid"], sample_colors["sample_color"]))
+            else:  # sample_name
+                sample_colors = (
+                    self.samples_df.filter(pl.col("sample_name").is_in(umap_df[id_col].unique()))
+                    .select(["sample_name", "sample_color"])
+                    .to_dict(as_series=False)
+                )
+                color_map = dict(zip(sample_colors["sample_name"], sample_colors["sample_color"]))
+
+            # Map colors into dataframe
+            umap_df["color"] = [color_map.get(x, "#1f77b4") for x in umap_df[id_col]]  # fallback to blue
+            # Update the ColumnDataSource with new color column
+            source = ColumnDataSource(umap_df)
+            scatter = p.scatter(
+                "UMAP1",
+                "UMAP2",
+                size=markersize,
+                alpha=alpha,
+                color="color",
+                source=source,
+            )
+        else:
+            scatter = p.scatter(
+                "UMAP1",
+                "UMAP2",
+                size=markersize,
+                alpha=alpha,
+                color="blue",
+                source=source,
+            )
+
+    # Create comprehensive hover tooltips with all sample information
+    tooltip_list = []
+
+    # Columns to exclude from tooltips (file paths and internal/plot fields)
+    excluded_cols = {"file_source", "file_path", "sample_path", "map_id", "UMAP1", "UMAP2", "ms1", "ms2", "size"}
+
+    # Add all sample dataframe columns to tooltips, skipping excluded ones
+    for col in samples_pd.columns:
+        if col in excluded_cols:
+            continue
+        if col in umap_df.columns:
+            if col == "sample_color":
+                # Display sample_color as a colored swatch
+                tooltip_list.append(("color", "$color[swatch]:sample_color"))
+            elif umap_df[col].dtype in ["float64", "float32"]:
+                tooltip_list.append((col, f"@{col}{{0.00}}"))
+            else:
+                tooltip_list.append((col, f"@{col}"))
+
+    hover = HoverTool(
+        tooltips=tooltip_list,
+        renderers=[scatter],
+    )
+    p.add_tools(hover)
+
+    # Add legend if using categorical coloring
+    if color_mapper and not isinstance(color_mapper, LinearColorMapper) and colorby:
+        # Only set legend properties if legends exist (avoid Bokeh warning when none created)
+        if getattr(p, "legend", None) and len(p.legend) > 0:
+            p.legend.location = "top_left"
+            p.legend.click_policy = "hide"
+
+    # Apply consistent save/display behavior
+    if filename is not None:
+        # Convert relative paths to absolute paths using study folder as base
+        import os
+        if not os.path.isabs(filename):
+            filename = os.path.join(self.folder, filename)
+
+        # Convert to absolute path for logging
+        abs_filename = os.path.abspath(filename)
+
+        # Use isolated file saving
+        _isolated_save_plot(p, filename, abs_filename, self.logger, "UMAP Plot")
+    else:
+        # Show in notebook when no filename provided
+        _isolated_show_notebook(p)
+    return p
+
+
 def plot_tic(
     self,
     samples=100,
@@ -2179,7 +2521,7 @@ def plot_tic(
     from bokeh.io.export import export_png
     from masster.study.helpers import get_tic
 
-    sample_uids = self._get_sample_uids(samples)
+    sample_uids = self._get_samples_uids(samples)
     if not sample_uids:
         self.logger.error("No valid sample_uids provided for TIC plotting.")
         return
@@ -2312,3 +2654,16 @@ def plot_tic(
         _isolated_show_notebook(p)
 
     return p
+
+
+def plot_pca(self, *args, **kwargs):
+    """Deprecated: Use plot_samples_pca instead."""
+    import warnings
+    warnings.warn("plot_pca is deprecated, use plot_samples_pca instead", DeprecationWarning, stacklevel=2)
+    return self.plot_samples_pca(*args, **kwargs)
+
+def plot_umap(self, *args, **kwargs):
+    """Deprecated: Use plot_samples_umap instead."""
+    import warnings
+    warnings.warn("plot_umap is deprecated, use plot_samples_umap instead", DeprecationWarning, stacklevel=2)
+    return self.plot_samples_umap(*args, **kwargs)

masster/study/processing.py

@@ -166,7 +166,7 @@ def align(self, **kwargs):
     # end of parameter initialization
 
     # Store parameters in the Study object
-    self.store_history(["align"], params.to_dict())
+    self.update_history(["align"], params.to_dict())
     self.logger.debug("Parameters stored to align")
 
     # Generate temporary feature maps on-demand from features_df instead of using cached data
@@ -370,7 +370,7 @@ def find_ms2(self, **kwargs):
     # end of parameter initialization
 
     # Store parameters in the Study object
-    self.store_history(["find_ms2"], params.to_dict())
+    self.update_history(["find_ms2"], params.to_dict())
     self.logger.debug("Parameters stored to find_ms2")
 
     data = []
@@ -551,7 +551,7 @@ def _integrate_chrom_impl(self, **kwargs):
     # end of parameter initialization
 
     # Store parameters in the Study object
-    self.store_history(["integrate_chrom"], params.to_dict())
+    self.update_history(["integrate_chrom"], params.to_dict())
     self.logger.debug("Parameters stored to integrate_chrom")
 
     # Get parameter values for use in the method
@@ -769,7 +769,7 @@ def integrate(self, **kwargs):
     # end of parameter initialization
 
     # Store parameters in the Study object
-    self.store_history(["integrate"], params.to_dict())
+    self.update_history(["integrate"], params.to_dict())
     self.logger.debug("Parameters stored to integrate")
 
     # Call the original integrate_chrom function with extracted parameters

masster/study/save.py

@@ -59,13 +59,16 @@ def save(self, filename=None, add_timestamp=True, compress=False):
 
     # Use compressed mode for large datasets
     if compress:
-        self._save_study5_compressed(filename)
+        from masster.study.h5 import _save_study5_compressed
+        _save_study5_compressed(self, filename)
     else:
-        self._save_study5(filename)
+        from masster.study.h5 import _save_study5
+        _save_study5(self, filename)
 
     if self.consensus_map is not None:
         # save the features as a separate file
-        self._save_consensusXML(filename=filename.replace(".study5", ".consensusXML"))
+        from masster.study.save import _save_consensusXML
+        _save_consensusXML(self, filename=filename.replace(".study5", ".consensusXML"))
     self.filename = filename
 
 
@@ -211,4 +214,5 @@ def save_consensus(self, **kwargs):
     if self.consensus_map is None:
         self.logger.error("No consensus map found.")
         return
-    self._save_consensusXML(**kwargs)
+    from masster.study.save import _save_consensusXML
+    _save_consensusXML(self, **kwargs)