Skip to content

Ground Truth Validation

glc.test_annotations(predictions, annot_df, annot_clf_col='lm_subclass', verbose=True)

Evaluate annotations against GLC predicted classes.

Parameters:

Name Type Description Default
predictions Dict[int, List[Tuple[str, float]]]

Dictionary mapping feature_id -> list of (class, score) tuples. Intended to be glc.GLCModel.predict_all() output.

 required
annot_df DataFrame

DataFrame with true annotations and must include a 'peak_id' column.

 required
annot_clf_col str

Column name containing the annotated class. (default: 'lm_subclass')

 'lm_subclass'
verbose bool

If True, prints warnings for missing features. These should be annotations not in the GGM main subgraph, which GLC cannot predict.

 True

Returns:

Type Description
DataFrame

A DataFrame comparing predictions to annotations.
Source code in src/glc/ground_truth_validation.py 
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
def test_annotations(
    predictions: Dict[int, List[Tuple[str, float]]],
    annot_df: pd.DataFrame,
    annot_clf_col: str = "lm_subclass",
    verbose: bool = True,
) -> pd.DataFrame:
    """
    Evaluate annotations against GLC predicted classes.

    Args:
        predictions: Dictionary mapping feature_id -> list of (class, score) tuples. Intended to be ```glc.GLCModel.predict_all()``` output.
        annot_df: DataFrame with true annotations and must include a 'peak_id' column.
        annot_clf_col: Column name containing the annotated class. (default: 'lm_subclass')
        verbose: If True, prints warnings for missing features. These should be annotations not in the GGM main subgraph, which GLC cannot predict.

    Returns:
        A DataFrame comparing predictions to annotations. 
    """

    results = []

    for feature, true_class in zip(
        annot_df["peak_id"].astype(int), annot_df[annot_clf_col]
    ):
        # Prediction retrieval
        pred_list = predictions.get(feature)

        if not pred_list:
            if verbose:
                print(
                    f"Feature {feature} skipped (missing or not in GGM main subgraph)."
                )
            continue

        # Top prediction
        top_class, top_score = pred_list[0]

        # Count equally highest scores
        n_top = sum(1 for _, s in pred_list if s == top_score)

        results.append(
            {
                "feature": feature,
                "annot_class": true_class,
                "pred_class": top_class,
                "pred_score": top_score,
                "n_top": n_top,
            }
        )

    # Build output DataFrame
    df = pd.DataFrame(results)
    df["correct"] = df["annot_class"] == df["pred_class"]

    return df

glc.PlotConfusionMatrices

Plot confusion matrices and compute precision/recall/F1 metrics of GLC predictions against ground truth annotations.

Source code in src/glc/ground_truth_validation.py 
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
class PlotConfusionMatrices:
    """
    Plot confusion matrices and compute precision/recall/F1 metrics of GLC predictions against ground truth annotations. 
    """

    def __init__(self, df: pd.DataFrame) -> None:
        self.df = df.copy()

    @staticmethod
    def _extract_codes(code: str) -> Tuple[Optional[str], Optional[str], Optional[str]]:
        if len(code) == 6:
            return code[:2], code[2:4], code[4:6]
        elif len(code) == 4:
            return code[:2], code[2:4], "00"
        else:
            return None, None, None

    def _add_misc_labels(self, df: pd.DataFrame, classes: List[str]) -> List[str]:
        preds: List[str] = []
        for _, row in df.iterrows():
            if row["pred_class"] in classes:
                preds.append(row["pred_class"])
            else:
                pred_main = row["pred_class"].split("[")[1][:4]
                annot_main = row["annot_class"].split("[")[1][:4]
                if pred_main == annot_main:
                    preds.append("Other: correct main class")
                else:
                    preds.append("Other: incorrect main class")
        return preds

    def _add_null_labels(self, df: pd.DataFrame, preds: List[str]) -> List[str]:
        updated_pred_axis_labels: List[str] = []
        for pred, pred_axis_label in zip(df["pred_class"], preds):
            if pred == "No prediction [000000]":
                updated_pred_axis_labels.append("No prediction")
            else:
                updated_pred_axis_labels.append(pred_axis_label)
        return updated_pred_axis_labels

    def _get_box_positions(self, class_labels: List[str], plot_all: bool) -> Tuple[List[int], List[int]]:
        count, delta = 0, 0
        running_code = ""
        start_pos: List[int] = []
        stop_pos: List[int] = []

        for lab in class_labels:
            try:
                code = lab.split("[")[1][2:4]
                if code != running_code:
                    running_code = code
                    start_pos.append(count - delta)
                    stop_pos.append(count)
                    delta = 0
            except Exception:
                start_pos.append(count - delta)
                stop_pos.append(count)
                break

            count += 1
            delta += 1

        if plot_all:
            start_pos.append(count - delta)
            stop_pos.append(count)

        return start_pos, stop_pos

    def _color_axes_ticks(self, axs: plt.Axes, annotation_classes: List[str], color: str = "#ec4743") -> None:
        for label in axs.get_xticklabels():
            if label.get_text() not in annotation_classes:
                label.set_color(color)

        for label in axs.get_yticklabels():
            if label.get_text() not in annotation_classes:
                label.set_color(color)

    def _get_cm_labels(self, plot_all: bool = True) -> List[str]:
        _df = self.df.copy()
        _df["pred_class"].fillna("No prediction [000000]", inplace=True)

        if plot_all:
            label_df = pd.DataFrame(pd.unique(_df[["annot_class", "pred_class"]].values.ravel()), columns=["class"])
        else:
            label_df = pd.DataFrame(pd.unique(_df["annot_class"]), columns=["class"])

        label_df["code"] = label_df["class"].str.extract(r"\[(.*?)\]")
        label_df[["category_code", "main_code", "sub_code"]] = label_df["code"].apply(
            lambda x: pd.Series(self._extract_codes(x))
        )
        label_df = label_df.sort_values(by=["category_code", "main_code", "sub_code"])

        class_labels = [i for i in label_df["class"] if i != "No prediction [000000]"]
        return class_labels

    def _plot_confusion_matrix(self, df: Optional[pd.DataFrame] = None, axs: Optional[plt.Axes] = None, vmax: int = 5):
        _df = (df.copy() if df is not None else self.df.copy())

        if axs is None:
            fig, axs = plt.subplots(figsize=(10, 8))

        class_labels = self._get_cm_labels(plot_all=True)

        preds = _df["pred_class"].tolist()
        if "No prediction" in preds:
            preds = self._add_null_labels(_df, preds)
            class_labels.extend(["No prediction"])

        cm = confusion_matrix(_df["annot_class"].tolist(), preds, labels=class_labels)

        sns.heatmap(
            cm, annot=True, fmt="d", cmap="Greens", vmax=vmax,
            xticklabels=class_labels, yticklabels=class_labels,
            linewidths=3, ax=axs
        )
        axs.set_xlabel("Predicted sub class", fontsize=12)
        axs.set_ylabel("True sub class", fontsize=12)

        annotation_classes = (
            df["annot_class"].unique().tolist()
            if df is not None else self.df["annot_class"].unique().tolist()
        )
        self._color_axes_ticks(axs, annotation_classes)

        start_pos, stop_pos = self._get_box_positions(class_labels, plot_all=True)

        square_color = "#0096FF"
        for start, stop in zip(start_pos, stop_pos):
            plt.plot([start, start], [start, stop], color=square_color, linewidth=2.5)
            plt.plot([start, stop], [start, start], color=square_color, linewidth=2.5)
            plt.plot([start, stop], [stop, stop], color=square_color, linewidth=2.5)
            plt.plot([stop, stop], [start, stop], color=square_color, linewidth=2.5)

        return axs, preds, class_labels

    def _plot_confusion_matrix_short(self, df=None, axs=None, vmax=5):
        _df = (df.copy() if df is not None else self.df.copy())

        if axs is None:
            fig, axs = plt.subplots(figsize=(10, 8))

        class_labels = self._get_cm_labels(plot_all=False)

        preds = self._add_misc_labels(_df, class_labels)
        class_labels.extend(["Other: correct main class", "Other: incorrect main class"])

        if "No prediction" in preds:
            preds = self._add_null_labels(_df, preds)
            class_labels.extend(["No prediction"])

        cm = confusion_matrix(_df["annot_class"].tolist(), preds, labels=class_labels)
        cm = cm[:-2, :]
        if "No prediction" in class_labels:
            cm = cm[:-1, :]

        y_class_labels = [
            i for i in class_labels
            if i not in ["Other: correct main class", "Other: incorrect main class", "No prediction"]
        ]

        sns.heatmap(
            cm, annot=True, fmt="d", cmap="Greens", vmax=vmax,
            xticklabels=class_labels, yticklabels=y_class_labels,
            linewidths=3, ax=axs
        )
        axs.set_xlabel("Predicted sub class", fontsize=12)
        axs.set_ylabel("True sub class", fontsize=12)

        start_pos, stop_pos = self._get_box_positions(class_labels, plot_all=False)
        square_color = "#0096FF"
        for start, stop in zip(start_pos, stop_pos):
            plt.plot([start, start], [start, stop], color=square_color, linewidth=2.5)
            plt.plot([start, stop], [start, start], color=square_color, linewidth=2.5)
            plt.plot([start, stop], [stop, stop], color=square_color, linewidth=2.5)
            plt.plot([stop, stop], [start, stop], color=square_color, linewidth=2.5)

        axs.axvline(x=stop_pos[-1] + 0.1, color="#A9A9A9", linewidth=2.5)

        return axs, preds, class_labels

    def plot_annotation_confusion_matrix(
        self,
        df: Optional[pd.DataFrame] = None,
        truncate: bool = True,
        vmax: int = 5,
        barplot: bool = False,
        barplot_title: str = "Number of sub classes that are top-ranking:",
    ) -> Tuple[plt.Figure, plt.Axes]:
        """Create a confusion matrix plot with optional top-k barplot.

        Args:
            df: Optional DataFrame to use. If None, uses evaluator's DataFrame copy.
            truncate: If True, use simplified confusion matrix. Only classes that are in the ground truth are shown.
            vmax: Max heatmap color value.
            barplot: Whether to create a proportional top-k stacked bar above the heatmap.
            barplot_title: Title for the barplot legend.

        Returns:
            Tuple of (figure, heatmap_axes).
        """
        _df = (df.copy() if df is not None else self.df.copy())

        if barplot:
            fig = plt.figure(figsize=(10, 10))
            gs = fig.add_gridspec(2, 1, height_ratios=[1, 6], hspace=0.2)
            ax_heatmap = fig.add_subplot(gs[1])
        else:
            fig, ax_heatmap = plt.subplots(figsize=(10, 8))

        if truncate:
            axs, preds, xtick_labels = self._plot_confusion_matrix_short(df=_df, axs=ax_heatmap, vmax=vmax)
        else:
            axs, preds, xtick_labels = self._plot_confusion_matrix(df=_df, axs=ax_heatmap, vmax=vmax)

        if barplot:
            _df["heatmap_col"] = preds

            grouped = _df.groupby(["heatmap_col", "n_top"]).size().unstack(fill_value=0)
            grouped = grouped.div(grouped.sum(axis=1), axis=0)

            new_idxs = list(set(xtick_labels) - set(grouped.index.tolist()))
            grouped = pd.concat([grouped, pd.DataFrame(index=new_idxs, columns=grouped.columns)])
            grouped = grouped.loc[xtick_labels]

            heatmap_pos = ax_heatmap.get_position()
            bar_pos = [heatmap_pos.x0, heatmap_pos.y1 + 0.01, heatmap_pos.width, 0.075]
            ax_bar = fig.add_axes(bar_pos)

            cmap = plt.get_cmap("Spectral_r")
            sns.set_style("darkgrid")
            grouped.plot(kind="bar", stacked=True, colormap=cmap, ax=ax_bar, width=1, legend=None)

            ax_bar.set_xticks([])
            ax_bar.set_xticklabels([])
            ax_bar.set_ylabel("Proportion")

            legend_patches = [
                mpatches.Patch(color=cmap((col - 1) / (max(max(grouped.columns) - 1, 1))), label=str(col))
                for col in grouped.columns
            ]

            fig.legend(handles=legend_patches, loc="upper left", ncol=len(grouped.columns),
                       bbox_to_anchor=(0.1, 0.85), frameon=False, title=barplot_title)

            plt.tight_layout()
            plt.show()

        return fig, axs

    def peformance_metrics(self, df: Optional[pd.DataFrame] = None, return_df: bool = False, level: str = "sub") -> Union[pd.DataFrame, dict]:
        """Compute classification metrics (precision/recall/f1) for the dataset.

        Args:
            df: Optional DataFrame to use. If None, uses evaluator's DataFrame copy.
            return_df: If True, return a pandas DataFrame; otherwise return a dict.
            level: 'sub' for subclass-level metrics or 'main' for main-class metrics.

        Returns:
            Either a pandas DataFrame (if return_df=True) or a dict with metrics.
        """
        _df = (df.copy() if df is not None else self.df.copy())

        _df["pred_class"].fillna("No prediction [000000]", inplace=True)

        if level == "sub":
            unique_labels = _df["annot_class"].unique()
            y_true = _df["annot_class"].values
            y_preds = _df["pred_class"].values
        elif level == "main":
            _df["main_annot_class"] = _df["annot_class"].apply(lambda x: x.split("[")[1][:4])
            _df["main_pred_class"] = _df["pred_class"].apply(lambda x: x.split("[")[1][:4])
            unique_labels = _df["main_annot_class"].unique()
            y_true = _df["main_annot_class"].values
            y_preds = _df["main_pred_class"].values
        else:
            raise ValueError("level must be 'sub' or 'main'")

        y_pred_filtered = np.where(np.isin(y_preds, unique_labels), y_preds, "other")

        report = classification_report(y_true, y_pred_filtered, labels=unique_labels, output_dict=True, zero_division=0)

        if return_df:
            return pd.DataFrame(report).T
        else:
            return report

peformance_metrics(df=None, return_df=False, level='sub')

Compute classification metrics (precision/recall/f1) for the dataset.

Parameters:

Name Type Description Default
df Optional[DataFrame]

Optional DataFrame to use. If None, uses evaluator's DataFrame copy.

 None
return_df bool

If True, return a pandas DataFrame; otherwise return a dict.

 False
level str

'sub' for subclass-level metrics or 'main' for main-class metrics.

 'sub'

Returns:

Type Description
Union[DataFrame, dict]

Either a pandas DataFrame (if return_df=True) or a dict with metrics.
Source code in src/glc/ground_truth_validation.py 
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
def peformance_metrics(self, df: Optional[pd.DataFrame] = None, return_df: bool = False, level: str = "sub") -> Union[pd.DataFrame, dict]:
    """Compute classification metrics (precision/recall/f1) for the dataset.

    Args:
        df: Optional DataFrame to use. If None, uses evaluator's DataFrame copy.
        return_df: If True, return a pandas DataFrame; otherwise return a dict.
        level: 'sub' for subclass-level metrics or 'main' for main-class metrics.

    Returns:
        Either a pandas DataFrame (if return_df=True) or a dict with metrics.
    """
    _df = (df.copy() if df is not None else self.df.copy())

    _df["pred_class"].fillna("No prediction [000000]", inplace=True)

    if level == "sub":
        unique_labels = _df["annot_class"].unique()
        y_true = _df["annot_class"].values
        y_preds = _df["pred_class"].values
    elif level == "main":
        _df["main_annot_class"] = _df["annot_class"].apply(lambda x: x.split("[")[1][:4])
        _df["main_pred_class"] = _df["pred_class"].apply(lambda x: x.split("[")[1][:4])
        unique_labels = _df["main_annot_class"].unique()
        y_true = _df["main_annot_class"].values
        y_preds = _df["main_pred_class"].values
    else:
        raise ValueError("level must be 'sub' or 'main'")

    y_pred_filtered = np.where(np.isin(y_preds, unique_labels), y_preds, "other")

    report = classification_report(y_true, y_pred_filtered, labels=unique_labels, output_dict=True, zero_division=0)

    if return_df:
        return pd.DataFrame(report).T
    else:
        return report

plot_annotation_confusion_matrix(df=None, truncate=True, vmax=5, barplot=False, barplot_title='Number of sub classes that are top-ranking:')

Create a confusion matrix plot with optional top-k barplot.

Parameters:

Name Type Description Default
df Optional[DataFrame]

Optional DataFrame to use. If None, uses evaluator's DataFrame copy.

 None
truncate bool

If True, use simplified confusion matrix. Only classes that are in the ground truth are shown.

 True
vmax int

Max heatmap color value.

 5
barplot bool

Whether to create a proportional top-k stacked bar above the heatmap.

 False
barplot_title str

Title for the barplot legend.

 'Number of sub classes that are top-ranking:'

Returns:

Type Description
Tuple[Figure, Axes]

Tuple of (figure, heatmap_axes).
Source code in src/glc/ground_truth_validation.py 
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
def plot_annotation_confusion_matrix(
    self,
    df: Optional[pd.DataFrame] = None,
    truncate: bool = True,
    vmax: int = 5,
    barplot: bool = False,
    barplot_title: str = "Number of sub classes that are top-ranking:",
) -> Tuple[plt.Figure, plt.Axes]:
    """Create a confusion matrix plot with optional top-k barplot.

    Args:
        df: Optional DataFrame to use. If None, uses evaluator's DataFrame copy.
        truncate: If True, use simplified confusion matrix. Only classes that are in the ground truth are shown.
        vmax: Max heatmap color value.
        barplot: Whether to create a proportional top-k stacked bar above the heatmap.
        barplot_title: Title for the barplot legend.

    Returns:
        Tuple of (figure, heatmap_axes).
    """
    _df = (df.copy() if df is not None else self.df.copy())

    if barplot:
        fig = plt.figure(figsize=(10, 10))
        gs = fig.add_gridspec(2, 1, height_ratios=[1, 6], hspace=0.2)
        ax_heatmap = fig.add_subplot(gs[1])
    else:
        fig, ax_heatmap = plt.subplots(figsize=(10, 8))

    if truncate:
        axs, preds, xtick_labels = self._plot_confusion_matrix_short(df=_df, axs=ax_heatmap, vmax=vmax)
    else:
        axs, preds, xtick_labels = self._plot_confusion_matrix(df=_df, axs=ax_heatmap, vmax=vmax)

    if barplot:
        _df["heatmap_col"] = preds

        grouped = _df.groupby(["heatmap_col", "n_top"]).size().unstack(fill_value=0)
        grouped = grouped.div(grouped.sum(axis=1), axis=0)

        new_idxs = list(set(xtick_labels) - set(grouped.index.tolist()))
        grouped = pd.concat([grouped, pd.DataFrame(index=new_idxs, columns=grouped.columns)])
        grouped = grouped.loc[xtick_labels]

        heatmap_pos = ax_heatmap.get_position()
        bar_pos = [heatmap_pos.x0, heatmap_pos.y1 + 0.01, heatmap_pos.width, 0.075]
        ax_bar = fig.add_axes(bar_pos)

        cmap = plt.get_cmap("Spectral_r")
        sns.set_style("darkgrid")
        grouped.plot(kind="bar", stacked=True, colormap=cmap, ax=ax_bar, width=1, legend=None)

        ax_bar.set_xticks([])
        ax_bar.set_xticklabels([])
        ax_bar.set_ylabel("Proportion")

        legend_patches = [
            mpatches.Patch(color=cmap((col - 1) / (max(max(grouped.columns) - 1, 1))), label=str(col))
            for col in grouped.columns
        ]

        fig.legend(handles=legend_patches, loc="upper left", ncol=len(grouped.columns),
                   bbox_to_anchor=(0.1, 0.85), frameon=False, title=barplot_title)

        plt.tight_layout()
        plt.show()

    return fig, axs