From 846211e62dda82005195add9b4f8073042a0d3d7 Mon Sep 17 00:00:00 2001
From: Swainlab <peter.swain@ed.ac.uk>
Date: Mon, 3 Jul 2023 11:37:45 +0100
Subject: [PATCH] unfinished mods to bud_metrioc and buddings
---
.../core/reshapers/bud_metric.py | 50 ++++++++++++++++++-
src/postprocessor/core/reshapers/buddings.py | 11 ++--
2 files changed, 55 insertions(+), 6 deletions(-)
diff --git a/src/postprocessor/core/reshapers/bud_metric.py b/src/postprocessor/core/reshapers/bud_metric.py
index ee239221..9e9e1f70 100644
--- a/src/postprocessor/core/reshapers/bud_metric.py
+++ b/src/postprocessor/core/reshapers/bud_metric.py
@@ -105,7 +105,7 @@ class BudMetric(LineageProcess):
return output_df
-def _combine_daughter_tracks(tracks: pd.DataFrame):
+def _combine_daughter_tracks_old(tracks: pd.DataFrame):
"""
Combine multiple time series of daughter cells into one time series.
@@ -141,3 +141,51 @@ def _combine_daughter_tracks(tracks: pd.DataFrame):
(combined_tracks == old) | (np.isnan(combined_tracks) & np.isnan(old))
).all(), "yikes"
return pd.Series(combined_tracks, index=tracks.columns)
+
+
+def _combine_daughter_tracks(tracks: pd.DataFrame):
+ """
+ Combine multiple time series of daughter cells into one time series.
+
+ Concatenate daughter values into one time series starting with the first
+ daughter and replacing later values with the values from the next daughter,
+ and so on.
+
+ Parameters
+ ----------
+ tracks: a Signal
+ Data for all daughters, which are distinguished by different cell_labels,
+ for a particular trap and mother_label.
+ """
+ # sort by daughter IDs
+ bud_df = tracks.sort_index(level="cell_label")
+ # remove multi-index
+ no_rows = len(bud_df)
+ bud_df.index = range(no_rows)
+ # find time point of first non-NaN data point of each row
+ init_tps = [
+ bud_df.iloc[irow].first_valid_index() for irow in range(no_rows)
+ ]
+ # sort so that earliest daughter is first
+ sorted_rows = np.argsort(init_tps)
+ init_tps = np.sort(init_tps)
+ # combine data for all daughters
+ combined_tracks = np.nan * np.ones(tracks.columns.size)
+ for j, jrow in enumerate(sorted_rows):
+ # over-write with next earliest daughter
+ combined_tracks[bud_df.columns.get_loc(init_tps[j]) :] = (
+ bud_df.iloc[jrow].loc[init_tps[j] :].values
+ )
+ # ## OLD
+ # # find which row of sorted_df has the daughter for each time point
+ # tp_fvt: pd.Series = bud_df.apply(lambda x: x.first_valid_index(), axis=0)
+ # # combine data for all daughters
+ # old = np.nan * np.ones(tracks.columns.size)
+ # for bud_row in np.unique(tp_fvt.dropna().values).astype(int):
+ # ilocs = np.where(tp_fvt.values == bud_row)[0]
+ # old[ilocs] = bud_df.values[bud_row, ilocs]
+ # assert (
+ # (combined_tracks == old) | (np.isnan(combined_tracks) & np.isnan(old))
+ # ).all(), "yikes"
+ # ###
+ return pd.Series(combined_tracks, index=tracks.columns)
diff --git a/src/postprocessor/core/reshapers/buddings.py b/src/postprocessor/core/reshapers/buddings.py
index ba9fe2fc..90785bce 100644
--- a/src/postprocessor/core/reshapers/buddings.py
+++ b/src/postprocessor/core/reshapers/buddings.py
@@ -37,7 +37,7 @@ class buddings(LineageProcess):
"""
Generate dataframe of budding events.
- Find daughters for mothers in a Signal for which we have lineage data.
+ Find daughters for those mothers in a Signal with lineage data.
Create a dataframe indicating the time each daughter first appears.
We use the data from Signal only to find when the daughters appear, by
@@ -66,14 +66,15 @@ class buddings(LineageProcess):
columns=signal.columns,
)
buddings.columns.names = ["timepoint"]
- # get time of first non-NaN value of signal for every mother using Pandas
+ # get time of first non-NaN value of signal for every cell using Pandas
fvi = signal.apply(lambda x: x.first_valid_index(), axis=1)
# fill the budding events
for trap_mother_id, daughters in traps_mothers.items():
+ trap_daughter_ids = [
+ i for i in product((trap_mother_id[0],), daughters)
+ ]
times_of_bud_appearance = fvi.loc[
- fvi.index.intersection(
- list(product((trap_mother_id[0],), daughters))
- )
+ fvi.index.intersection(trap_daughter_ids)
].values
# ignore zeros - ignore buds in first image
daughters_idx = set(times_of_bud_appearance).difference({0})
--
GitLab