How can I calculate sequences of contact events?
Question:
I have a dataset that represents contact events between tumors. The dataset is grouped by the "base-cell" and then sorted on "Neighbor-cell" and "Time-frame", it looks like this:
index
base-cell
neighbor-cell
timeframe
0
Track_1
Track_4
1
1
Track_1
Track_4
2
2
Track_1
Track_4
3
3
Track_1
Track_4
4
4
Track_1
Track_4
8
5
Track_1
Track_4
9
6
Track_1
Track_4
10
7
Track_1
Track_6
1
8
Track_1
Track_6
2
Because the dataframe is grouped on base-tumor, I have multiple dataframes with ascending base-tumor.
The end result that I’m trying to get to is a dictionary with all tracks that contains a dictionary with all tracks that have a contact event with it, and then they contain a list of the frames where there is a sequence of contact events. It looks like this:
{Track_1: {Track_4: [[1,4], [8,10],
Track_6: [[1,2]]},
Track_2: {Track_5: [[10, 14], [20, 25], [28, 31]}}
What I’ve done till now is, I made an extra column that shows a 1 if there is a sequence and a 0 if there is no sequence of contact events.
def get_sequence(df):
for id, grp in df:
prev_id = grp['id_2'].shift(1).fillna(0)
prev_frame = grp['FRAME'].shift(1)
conditions = [
((grp['id_2'] == prev_id) &
(grp['FRAME']) - prev_frame == 1)
]
choises = [1]
grp['sequence'] = np.select(conditions, choises, default=0)
print(grp)
Now I’m stuck and don’t know if I’m going in the right direction and if so, how to take the next step.
Answers:
Here would be one way:
# Identify continuous timeframes.
df['consec'] = df.groupby(['base-tumor', 'neighbor-tumor'])['timeframe'].transform(lambda s: s.diff().ne(1).cumsum())
# Get timeframe intervals.
t_df = (df.groupby(['base-tumor', 'neighbor-tumor', 'consec']).
agg(t_start=('timeframe', 'first'), t_end=('timeframe', 'last')).
droplevel(-1)
)
t_df = t_df[t_df['t_start'].ne(t_df['t_end'])]
t_df['interval'] = list(zip(t_df['t_start'], t_df['t_end']))
# Convert to dictionary.
result = {k: g.droplevel(0)['interval'].groupby(level=0).agg(list).to_dict()
for k, g in t_df.drop(columns=['t_start', 't_end']).groupby(level=0)}
print(result)
{'Track_1': {'Track_4': [(1, 4), (8, 10)], 'Track_6': [(1, 2)]}}
I have a dataset that represents contact events between tumors. The dataset is grouped by the "base-cell" and then sorted on "Neighbor-cell" and "Time-frame", it looks like this:
index | base-cell | neighbor-cell | timeframe |
---|---|---|---|
0 | Track_1 | Track_4 | 1 |
1 | Track_1 | Track_4 | 2 |
2 | Track_1 | Track_4 | 3 |
3 | Track_1 | Track_4 | 4 |
4 | Track_1 | Track_4 | 8 |
5 | Track_1 | Track_4 | 9 |
6 | Track_1 | Track_4 | 10 |
7 | Track_1 | Track_6 | 1 |
8 | Track_1 | Track_6 | 2 |
Because the dataframe is grouped on base-tumor, I have multiple dataframes with ascending base-tumor.
The end result that I’m trying to get to is a dictionary with all tracks that contains a dictionary with all tracks that have a contact event with it, and then they contain a list of the frames where there is a sequence of contact events. It looks like this:
{Track_1: {Track_4: [[1,4], [8,10],
Track_6: [[1,2]]},
Track_2: {Track_5: [[10, 14], [20, 25], [28, 31]}}
What I’ve done till now is, I made an extra column that shows a 1 if there is a sequence and a 0 if there is no sequence of contact events.
def get_sequence(df):
for id, grp in df:
prev_id = grp['id_2'].shift(1).fillna(0)
prev_frame = grp['FRAME'].shift(1)
conditions = [
((grp['id_2'] == prev_id) &
(grp['FRAME']) - prev_frame == 1)
]
choises = [1]
grp['sequence'] = np.select(conditions, choises, default=0)
print(grp)
Now I’m stuck and don’t know if I’m going in the right direction and if so, how to take the next step.
Here would be one way:
# Identify continuous timeframes.
df['consec'] = df.groupby(['base-tumor', 'neighbor-tumor'])['timeframe'].transform(lambda s: s.diff().ne(1).cumsum())
# Get timeframe intervals.
t_df = (df.groupby(['base-tumor', 'neighbor-tumor', 'consec']).
agg(t_start=('timeframe', 'first'), t_end=('timeframe', 'last')).
droplevel(-1)
)
t_df = t_df[t_df['t_start'].ne(t_df['t_end'])]
t_df['interval'] = list(zip(t_df['t_start'], t_df['t_end']))
# Convert to dictionary.
result = {k: g.droplevel(0)['interval'].groupby(level=0).agg(list).to_dict()
for k, g in t_df.drop(columns=['t_start', 't_end']).groupby(level=0)}
print(result)
{'Track_1': {'Track_4': [(1, 4), (8, 10)], 'Track_6': [(1, 2)]}}