tables_from_markdown

In [2]:

from great_tables import GT, md, html, style, loc, px
import polars as pl
import mdpd
from zfish.plot.plot_commons import cmaps

MONOFONT = "source code pro"
TABLEFONT = None


def gt_render_pdf(
    df,
    header="Base Dimensions",
    hide=None,
    font=MONOFONT,
    weith="normal",
    size=px(14),
    columns=(),
):
    gt_df = (
        GT(df)
        .tab_style(
            style=style.text(
                color=cmaps.spatial_element_dark[3],
                font=MONOFONT,
                weight="normal",
                size=size,
            ),
            locations=loc.body(columns=list(columns)),
        )
        .tab_header(
            title=md(header),
        )
    )
    if hide is not None:
        return gt_df.cols_hide(hide)
    return gt_df


def gt_render(
    df,
    header="Base Dimensions",
    hide=None,
    font=MONOFONT,
    weith="normal",
    size=px(14),
    columns=(),
    v_padding=0.8,
    h_padding=0.8,
):
    gt_df = (
        GT(df)
        .tab_style(
            style=style.text(
                color=cmaps.spatial_element_dark[3],
                font=MONOFONT,
                weight="normal",
                size=size,
            ),
            locations=loc.body(columns=list(columns)),
        )
        .tab_header(
            title=md(header),
        )
        .opt_vertical_padding(scale=v_padding)
        .opt_horizontal_padding(scale=h_padding)
    )
    if hide is not None:
        return gt_df.cols_hide(hide)
    return gt_df


def gt_render_from_markdown(
    df,
    header="Base Dimensions",
    hide=None,
    size=px(14),
    columns=(),
    v_padding=0.8,
    h_padding=0.8,
):
    gt_df = (
        GT(df)
        .tab_header(
            title=md(header),
        )
        .opt_vertical_padding(scale=v_padding)
        .opt_horizontal_padding(scale=h_padding)
    )
    if hide is not None:
        return gt_df.cols_hide(hide)
    return gt_df.fmt_markdown(list(columns))

In [3]:

import polars as pl
import mdpd

df_data_structures = pl.DataFrame(
    mdpd.from_md("""
| Data Structure        | Description |
|------------------------|-----------------------------------------------------------|
| **Raster Data**        |  |
| Image                  | Fluorescence intensity measurements sampled on a regular grid. |
| Binary Mask            | Partition of an image into foreground and background, representing a single object. |
| Label Image            | Partitioning of an image into *n* objects with integer IDs (labels). `0` for background. Represents a collection of objects. |
| Distance Transform     | Image with pixel values storing the distance to a binary mask. |
| Multiscale Pyramid     | Collection of images sampled at different scales (pyramid schedule) with shared coverage. |
| **Vector Data**        |  |
| Point / Vertex         | Position in space, potentially associated with an object or features: `(z, y, x)`. |
| Vector                 | *Difference* between two points, or direction and magnitude: `(dz, dy, dx)`. |
| Line Segment / Edge    | 1D path between two points (also 1-simplex): `(z, y, x), (z, y, x)` *or* `(z, y, x), (dz, dy, dx)`. |
| Face                   | Simple 2D surface, *often* a triangle. Defined by points and their connectivity. |
| Mesh                   | Representation of objects or surfaces using points, edges, and faces placed in physical space. |
| Graph (Spatial)        | Mesh with only points and edges to represent objects in space and their relationships. |
| Neighborhood           | Spatial graph representing object neighborhoods, for example using sparse arrays. |
| **Non-Spatial**\*      |  |
| Feature Table          | Quantitative or categorical features (columns) associated with a set of elements (rows). |
| Graph                  | Nodes and edges representing entities and relationships between them. Many possible representations, *usually* sparse. |
| Hierarchy Table        | Graph encoding relationships (rows) between pairs of objects (columns; 2 IDs). Edge list representation of a graph. |
""")
)


def gt_bold_column_labels(df):
    return (
        GT(df_data_structures)
        .fmt_markdown()
        .cols_label(
            **{col: md(f"**{col}**") for col in df.columns},
        )
    )

In [4]:

gt_bold_column_labels(df_data_structures).tab_source_note(
    source_note=md("*Since all our measurements are derived from pixels they can *always* be associated with spatial objects. But we often drop the spatial data components, like when doing regression modelling with extracted features, and think of them as non-spatial.")
)

Data Structure	Description
Raster Data
Image	Fluorescence intensity measurements sampled on a regular grid.
Binary Mask	Partition of an image into foreground and background, representing a single object.
Label Image	Partitioning of an image into n objects with integer IDs (labels). `0` for background. Represents a collection of objects.
Distance Transform	Image with pixel values storing the distance to a binary mask.
Multiscale Pyramid	Collection of images sampled at different scales (pyramid schedule) with shared coverage.
Vector Data
Point / Vertex	Position in space, potentially associated with an object or features: `(z, y, x)`.
Vector	Difference between two points, or direction and magnitude: `(dz, dy, dx)`.
Line Segment / Edge	1D path between two points (also 1-simplex): `(z, y, x), (z, y, x)` or `(z, y, x), (dz, dy, dx)`.
Face	Simple 2D surface, often a triangle. Defined by points and their connectivity.
Mesh	Representation of objects or surfaces using points, edges, and faces placed in physical space.
Graph (Spatial)	Mesh with only points and edges to represent objects in space and their relationships.
Neighborhood	Spatial graph representing object neighborhoods, for example using sparse arrays.
Non-Spatial*
Feature Table	Quantitative or categorical features (columns) associated with a set of elements (rows).
Graph	Nodes and edges representing entities and relationships between them. Many possible representations, usually sparse.
Hierarchy Table	Graph encoding relationships (rows) between pairs of objects (columns; 2 IDs). Edge list representation of a graph.
Since all our measurements are derived from pixels they can always* be associated with spatial objects. But we often drop the spatial data components, like when doing regression modelling with extracted features, and think of them as non-spatial.

In [5]:

df_dims_base = pl.DataFrame(
    mdpd.from_md("""
| Type | Dimension | Coordinate | Coordinate Description |
| :---|:--|:-------------|:------------|
| nominal | `roi` | `['site0', 'site1', ..., 'site211']` | Named, bounded regions of space. |
| temporal | `t`  | `[0.0, 15.0, 30.0, ..., 1000.0] s` | Acquisition timestamps. |
| spatial | `z`  | `[0.0, 271.0] μm` | Bounded continuous z-coordinate. |
| spatial | `y`  | `[0.0, 650.0] μm` | Bounded continuous y-coordinate. |
| spatial | `x`  | `[0.0, 650.0] μm` | Bounded continuous x-coordinate. |
| ordinal | `m`   | `[0, 1, 2, 3, 4, 5]` | Sampling scales of the pyramid grids. |
| nominal | `c`   | `['DAPI.0', 'PCNA.0', ..., 'Nanog.3']` | Molecular targets with acquisition cycle. |
| ratio   | `val` | `[0.0, 4.2, 5.4, ...] AU` | Intensity histogram for a channel. |
| nominal | `o`   | `['embryo', 'cell', ..., 'nuc']` | Types of segmented objects. |
| nominal | `lbl`* | `[1, 2, ... n]` | Instance labels of an object type. |
""")
)

In [6]:

tbl = (
    gt_render_from_markdown(df_dims_base, columns=("Dimension", "Coordinate"))
    .cols_move_to_start("Dimension")
    .cols_label({"Dimension": "Dim"})
)
tbl

Base Dimensions
Dim	Type	Coordinate	Coordinate Description
`roi`	nominal	`['site0', 'site1', ..., 'site211']`	Named, bounded regions of space.
`t`	temporal	`[0.0, 15.0, 30.0, ..., 1000.0] s`	Acquisition timestamps.
`z`	spatial	`[0.0, 271.0] μm`	Bounded continuous z-coordinate.
`y`	spatial	`[0.0, 650.0] μm`	Bounded continuous y-coordinate.
`x`	spatial	`[0.0, 650.0] μm`	Bounded continuous x-coordinate.
`m`	ordinal	`[0, 1, 2, 3, 4, 5]`	Sampling scales of the pyramid grids.
`c`	nominal	`['DAPI.0', 'PCNA.0', ..., 'Nanog.3']`	Molecular targets with acquisition cycle.
`val`	ratio	`[0.0, 4.2, 5.4, ...] AU`	Intensity histogram for a channel.
`o`	nominal	`['embryo', 'cell', ..., 'nuc']`	Types of segmented objects.
`lbl`*	nominal	`[1, 2, ... n]`	Instance labels of an object type.

In [7]:

df_dims_compound = pl.DataFrame(
    mdpd.from_md("""
| Type        | Dimension      | Components         | Description                                         |
|:------------|:---------------|:-------------------|:----------------------------------------------------|
| categorical | `image`        | `roi`, `m`, `c`    | Single-channel, single-scale image.                 |
| categorical | `label_image`  | `roi`, `m`, `o`    | Single-scale label image.                           |
| categorical | `label_object` | `roi`, `o`, `lbl`  | Segmented object (ID is valid across `m`).          |

""")
)

In [8]:

gt_render_from_markdown(
    df_dims_compound.drop('Type'),
    # hide="Type",
    header="Compound Dimensions",
    columns=("Dimension", "Components"),
    h_padding=1.8,
).cols_move_to_start("Dimension")

Compound Dimensions
Dimension	Components	Description
`image`	`roi`, `m`, `c`	Single-channel, single-scale image.
`label_image`	`roi`, `m`, `o`	Single-scale label image.
`label_object`	`roi`, `o`, `lbl`	Segmented object (ID is valid across `m`).

In [9]:

df_dims_slice = pl.DataFrame(
    mdpd.from_md("""
| Dimension           | Components                                 | Description                                      |
|---------------------|--------------------------------------------|--------------------------------------------------|
| `mc_image`          | `roi`, `m`, `c:n`                          | Multi-channel, single-scale image.               |
| `image_pyramid`     | `roi`, `m:n`, `c`                          | Single-channel, multi-scale image.               |
| `mc_image_pyramid`  | `roi`, `m:n`, `c:n`                        | Multi-channel, multi-scale image.                |
| `sub_label_image`   | `roi`, `m`, `o`, `lbl:n`                   | Label image formed from a subset of labels.      |
| `voxel_region`      | `roi`, `m`, `iz`, `iy`, `ix`               | Voxel without associated data.                   |
| `mc_mo_voxel`       | `roi`, `m`, `iz`, `iy`, `ix`, `c:n`, `o:n` | Voxel with channel data and object type data.    |
| `roi`               | `roi`, `m:n`, `c:n`, `o:n`                 | Region of interest images and label images.      |
| `dataset`           | `roi:n`, `m:n`, `c:n` `o:n`                | Dataset containing a collection of ROIs.         |
""")
)

df_dims_slice

shape: (8, 3)

Dimension	Components	Description
str	str	str
"`mc_image`"	"`roi`, `m`, `c:n`"	"Multi-channel, single-scale im…
"`image_pyramid`"	"`roi`, `m:n`, `c`"	"Single-channel, multi-scale im…
"`mc_image_pyramid`"	"`roi`, `m:n`, `c:n`"	"Multi-channel, multi-scale ima…
"`sub_label_image`"	"`roi`, `m`, `o`, `lbl:n`"	"Label image formed from a subs…
"`voxel_region`"	"`roi`, `m`, `iz`, `iy`, `ix`"	"Voxel without associated data."
"`mc_mo_voxel`"	"`roi`, `m`, `iz`, `iy`, `ix`, …	"Voxel with channel data and ob…
"`roi`"	"`roi`, `m:n`, `c:n`, `o:n`"	"Region of interest images and …
"`dataset`"	"`roi:n`, `m:n`, `c:n` `o:n`"	"Dataset containing a collectio…

In [10]:

gt_render_from_markdown(
    df_dims_slice,
    header="Slice Dimensions",
    columns=["Dimension", "Components"],
    h_padding=0.8,
).cols_move_to_start("Dimension")

Slice Dimensions
Dimension	Components	Description
`mc_image`	`roi`, `m`, `c:n`	Multi-channel, single-scale image.
`image_pyramid`	`roi`, `m:n`, `c`	Single-channel, multi-scale image.
`mc_image_pyramid`	`roi`, `m:n`, `c:n`	Multi-channel, multi-scale image.
`sub_label_image`	`roi`, `m`, `o`, `lbl:n`	Label image formed from a subset of labels.
`voxel_region`	`roi`, `m`, `iz`, `iy`, `ix`	Voxel without associated data.
`mc_mo_voxel`	`roi`, `m`, `iz`, `iy`, `ix`, `c:n`, `o:n`	Voxel with channel data and object type data.
`roi`	`roi`, `m:n`, `c:n`, `o:n`	Region of interest images and label images.
`dataset`	`roi:n`, `m:n`, `c:n` `o:n`	Dataset containing a collection of ROIs.

In [11]:

df_features_overview = pl.DataFrame(
    mdpd.from_md("""
| Feature Type       | Object ID                     | Resources                                  | Features         | Examples                                         |
|--------------------|-------------------------------|--------------------------------------------|------------------|--------------------------------------------------|
| Label              | `roi`, `o`, `lbl`             | `m`                                        | `f:label`        | Volume, Roundness, Principal Axes, Centroid Position |
| Intensity          | `roi`, `o`, `lbl`             | `m`, `c`                                   | `f:intensity`    | DAPI Abundance*, Concentration**, Variance       |
| Colocalization     | `roi`, `o`, `lbl`             | `m`, `c.0`, `c.1`                          | `f:coloc`        | DAPI Correlation                                 |
| Distance           | `roi`, `o`, `lbl`             | `m`, `o.to`, `[lbl.to]`, `dist_transf`     | `f:distance`     | Centroid Distance To Embryo Border               |
| Hierarchy (1:1)    | `roi`, `o`, `lbl`             | `o.from`, `f:in`                           | `f:in`           | Nuclear DAPI Abundance in Cell                   |
| Hierarchy (m:1)    | `roi`, `o`, `lbl`             | `o.from`, `agg_func`, `[f:in]`             | `Count`, `f:in`  | Mean Nuclear Volume in Embryo                    |
| Neighborhood       | `roi`, `o`, `lbl`             | `nhd_query`, `agg_func`, `[f:in]`          | `Count`, `f:in`  | Cell Count in Radius 30 μm Neighborhood          |
| Embedding          | `roi`, `o`, `lbl`             | `model`, `f:in`                            | `f:out`          | UMAP Embedding, Cell Cycle Phase, Cell Type      |
""") 
)
df_features_overview#.pipe(GT).fmt_markdown().tab_source_note(source_note="*Sum / **Mean intensity")

shape: (8, 5)

Feature Type	Object ID	Resources	Features	Examples
str	str	str	str	str
"Label"	"`roi`, `o`, `lbl`"	"`m`"	"`f:label`"	"Volume, Roundness, Principal A…
"Intensity"	"`roi`, `o`, `lbl`"	"`m`, `c`"	"`f:intensity`"	"DAPI Abundance*, Concentration…
"Colocalization"	"`roi`, `o`, `lbl`"	"`m`, `c.0`, `c.1`"	"`f:coloc`"	"DAPI Correlation"
"Distance"	"`roi`, `o`, `lbl`"	"`m`, `o.to`, `[lbl.to]`, `dist…	"`f:distance`"	"Centroid Distance To Embryo Bo…
"Hierarchy (1:1)"	"`roi`, `o`, `lbl`"	"`o.from`, `f:in`"	"`f:in`"	"Nuclear DAPI Abundance in Cell"
"Hierarchy (m:1)"	"`roi`, `o`, `lbl`"	"`o.from`, `agg_func`, `[f:in]`"	"`Count`, `f:in`"	"Mean Nuclear Volume in Embryo"
"Neighborhood"	"`roi`, `o`, `lbl`"	"`nhd_query`, `agg_func`, `[f:i…	"`Count`, `f:in`"	"Cell Count in Radius 30 μm Nei…
"Embedding"	"`roi`, `o`, `lbl`"	"`model`, `f:in`"	"`f:out`"	"UMAP Embedding, Cell Cycle Pha…

In [12]:

gt_render_from_markdown(
    df_features_overview,
    header="Feature Tables",
    columns=("Object ID", "Resources", "Features"),
    # h_padding=0.8,
).tab_source_note(source_note="*Sum / **Mean intensity")

Feature Tables
Feature Type	Object ID	Resources	Features	Examples
Label	`roi`, `o`, `lbl`	`m`	`f:label`	Volume, Roundness, Principal Axes, Centroid Position
Intensity	`roi`, `o`, `lbl`	`m`, `c`	`f:intensity`	DAPI Abundance, Concentration*, Variance
Colocalization	`roi`, `o`, `lbl`	`m`, `c.0`, `c.1`	`f:coloc`	DAPI Correlation
Distance	`roi`, `o`, `lbl`	`m`, `o.to`, `[lbl.to]`, `dist_transf`	`f:distance`	Centroid Distance To Embryo Border
Hierarchy (1:1)	`roi`, `o`, `lbl`	`o.from`, `f:in`	`f:in`	Nuclear DAPI Abundance in Cell
Hierarchy (m:1)	`roi`, `o`, `lbl`	`o.from`, `agg_func`, `[f:in]`	`Count`, `f:in`	Mean Nuclear Volume in Embryo
Neighborhood	`roi`, `o`, `lbl`	`nhd_query`, `agg_func`, `[f:in]`	`Count`, `f:in`	Cell Count in Radius 30 μm Neighborhood
Embedding	`roi`, `o`, `lbl`	`model`, `f:in`	`f:out`	UMAP Embedding, Cell Cycle Phase, Cell Type
Sum / *Mean intensity

In [13]:

FEATURE_SETS = {
    "ccp": ("CCP",),
    "ccp-feats": (
        "PCNA",
        "DAPI",
        "n_Volume",
        "n.cy_VolumeRatio",
        "n_Roundness",
        "n_Flatness",
    ),
    "zga": ("Pou5", "Nanog", "H3K27Ac", "H2B"),
    "s5p": ("Pol2.S5P",),
}
(
    GT(
        pl.DataFrame(
            [("", ["Pol2.S2P"])]
            + list(FEATURE_SETS.items())
            + [
                ("ccp-spline", ["bs(CCP, 3, fit_intercept=False)"]),
                ("ccp-asympt", ["CCP(a, b, c)"]),
                ("ccp-power", ["CCP(a, b)"]),
            ],
            orient="row",
            schema=["Component", "Features"],
        )
        .with_columns(
            pl.Series(
                "Description",
                [
                    "Pol-II-S2P (elongating)",
                    "Interphase CCP",
                    "Cell cycle phase features*",
                    "ZGA factors",
                    "Pol-II-S5P (initiating)",
                    "B-spline encoded interphase CCP (3 knots w/o intercept)",
                    "Asymptotic regression on CCP",
                    "Power curve regression on CCP",
                ],
            ),
        )
        .with_columns(pl.col("Features").list.join(", "))
        .with_columns(
            pl.Series(
                "category",
                [
                    "regression target",
                    "linear predictors",
                    "linear predictors",
                    "linear predictors",
                    "linear predictors",
                    "non-linear predictors",
                    "non-linear predictors",
                    "non-linear predictors",
                ],
            )
        )
    )
    .tab_source_note(
        source_note="*a subset of the features that went into the CCP model."
    )
    .tab_stub(rowname_col="Component", groupname_col="category")
    .tab_stubhead("Component")
    .tab_style(
            style=style.text(
                color=cmaps.spatial_element_dark[3],
                font=MONOFONT,
                weight="normal",
            ),
            locations=loc.body(columns=["Component"]),
        )
    .tab_options(column_labels_font_weight="bold")
    # .opt_stylize(style=2, color='blue')
)

Component	Features	Description
regression target
	Pol2.S2P	Pol-II-S2P (elongating)
linear predictors
ccp	CCP	Interphase CCP
ccp-feats	PCNA, DAPI, n_Volume, n.cy_VolumeRatio, n_Roundness, n_Flatness	Cell cycle phase features*
zga	Pou5, Nanog, H3K27Ac, H2B	ZGA factors
s5p	Pol2.S5P	Pol-II-S5P (initiating)
non-linear predictors
ccp-spline	bs(CCP, 3, fit_intercept=False)	B-spline encoded interphase CCP (3 knots w/o intercept)
ccp-asympt	CCP(a, b, c)	Asymptotic regression on CCP
ccp-power	CCP(a, b)	Power curve regression on CCP
*a subset of the features that went into the CCP model.

In [14]:

import polars as pl
from great_tables import GT

df_ccp_components = pl.DataFrame(
    {
        "Component": [
            "`PolarsSelector`",
            "`StratifiedTransformer`",
            "`StratifiedEmbedderAligned`",
            "`CcpTransformer`",
        ],
        "Description": [
            "Selects columns from a Polars DataFrame using a column selector.",
            "Meta-transformer that maps a transformer across subsets of the data by fitting one for each subset.",
            "Meta-transformer that maps an embedding (e.g., UMAP) on subsets of data and aligns the results using a rigid transformation.",
            "Cell cycle phase transformer that takes a pre-embedding as input, fits a principal circle, and returns `CCP`, `NormalizedCCP`, and `DistanceToCircleCCP`."
        ],
    }
)


gt_render_from_markdown(
    df_ccp_components,
    header="Scikit-learn Compatible CCP Components",
    columns=("Component", "Description"),
    h_padding=0.8,
)

Scikit-learn Compatible CCP Components
Component	Description
`PolarsSelector`	Selects columns from a Polars DataFrame using a column selector.
`StratifiedTransformer`	Meta-transformer that maps a transformer across subsets of the data by fitting one for each subset.
`StratifiedEmbedderAligned`	Meta-transformer that maps an embedding (e.g., UMAP) on subsets of data and aligns the results using a rigid transformation.
`CcpTransformer`	Cell cycle phase transformer that takes a pre-embedding as input, fits a principal circle, and returns `CCP`, `NormalizedCCP`, and `DistanceToCircleCCP`.

In [22]:

df_ab_stain_conditions

shape: (12, 9)

acquisition	wavelength	target_primary	host_primary	dilution_primary	secondary_is_fab	name_secondary	dilution_secondary	stain_type
i64	i64	str	str	str	bool	str	str	str
0	488	"Pol-II-S2P"	"Rat"	"1:250"	false	"d@rt 488"	"1:200"	"Antibody"
0	568	"FLAG"	"Mouse"	"1:500"	false	"d@ms 568"	"1:200"	"Antibody"
0	647	"PCNA"	"Rabbit"	"1:250"	false	"d@rb 647"	"1:200"	"Antibody"
1	488	"H3K27Ac"	"Rabbit"	"1:500"	true	"Zenon Rb 488"	"1:100"	"Antibody"
1	568	"bCatenin"	"Mouse"	"1:250"	false	"d@ms 568"	"1:400"	"Antibody"
…	…	…	…	…	…	…	…	…
2	568	"ALYREF"	"Mouse"	"1:250"	true	"Zenon Ms 555"	"1:100"	"Antibody"
2	647	"Pol-II-S5P"	"Rat"	"1:250"	false	"d@rt 647"	"1:200"	"Antibody"
3	488	"Nanog"	"Rabbit"	"1:280"	true	"Zenon Rb 488"	"1:28"	"Antibody"
3	568	"YAP"	"Mouse"	"1:50"	false	"d@ms 568"	"1:400"	"Antibody"
3	647	"XRN2"	"Rabbit"	"1:100"	true	"Zenon Rb 647"	"1:20"	"Antibody"

In [1]:

df_ab_stain_conditions = pl.read_csv("../Data/AB_stainings.csv")

df_ab_stain_conditions.select(
    pl.col("target_primary").alias("primary"),
    pl.col("host_primary").alias("host"),
    pl.col("dilution_primary").alias("dilution"),
    pl.col('name_secondary').alias('secondary'),
    pl.col('dilution_secondary').alias('secondary dilution'),
    pl.when(pl.col('secondary_is_fab')).then(pl.lit('IF direct')).otherwise(pl.lit('IF indirect')).alias('stain type'),

).pipe(GT)

primary	host	dilution	stain type	secondary name	secondary dilution
Pol-II-S2P	Rat	1:250	IF indirect	d@rt 488	1:200
FLAG	Mouse	1:500	IF indirect	d@ms 568	1:200
PCNA	Rabbit	1:250	IF indirect	d@rb 647	1:200
H3K27Ac	Rabbit	1:500	IF direct	Zenon Rb 488	1:100
bCatenin	Mouse	1:250	IF indirect	d@ms 568	1:400
pH3	Rabbit	1:150	IF direct	Zenon Rb 647	1:150
H2B	Rabbit	1:100	IF direct	Zenon Rb 488	1:250
ALYREF	Mouse	1:250	IF direct	Zenon Ms 555	1:100
Pol-II-S5P	Rat	1:250	IF indirect	d@rt 647	1:200
Nanog	Rabbit	1:280	IF direct	Zenon Rb 488	1:28
YAP	Mouse	1:50	IF indirect	d@ms 568	1:400
XRN2	Rabbit	1:100	IF direct	Zenon Rb 647	1:20

tables_from_markdown.ipynb