Basic usage with python
The following steps show how to get started with wellmap
in python:
Install
wellmap
from PyPI. Note that python≥3.6 is required:$ pip install wellmap
Write a TOML file describing the layout of an experiment. For example, the following layout might be used for a standard curve:
# The [row] and [col] sections specify which conditions are being tested in # which wells. The fields within these sections (e.g. `dilution`, `replicate`) # can be anything. If your plates aren't organized by row and column, there # are other ways to define the plate layout; see the "File format" section for # more details. [col] 1.dilution = 1e5 2.dilution = 1e4 3.dilution = 1e3 4.dilution = 1e2 5.dilution = 1e1 6.dilution = 1e0 [row] A.replicate = 1 B.replicate = 2 C.replicate = 3
Confirm that the layout is correct by using the
wellmap
command-line program to produce a visualization of the layout. This is an important step, because it’s much easier to spot mistakes in the visualization than in the layout file itself.$ wellmap std_curve.toml
This map shows that:
Each row is a different replicate.
Each column is a different dilution.
It is also possible to create maps like this directly from python, which may be useful in interactive sessions such as Jupyter notebooks:
>>> import wellmap >>> wellmap.show("std_curve.toml") <Figure size ... with 4 Axes>
Load the data from the experiment in question into a tidy data frame. Tidy data are easier to work with in general, and are required by
wellmap
in particular. If you aren’t familiar with the concept of tidy data, this article is a good introduction. The basic idea is to ensure that:Each variable is represented by a single column.
Each observation is represented by a single row.
If possible, it’s best to export data from the instrument that collected it directly to a tidy format. When this isn’t possible, though, you’ll need to tidy the data yourself. For example, consider the following data (which corresponds to the layout from above). This is qPCR data, where a higher \(C_q\) value indicates that less material is present. The data are shaped like the plate itself, e.g. a row in the data for every row on the plate, and a column in the data for every column on the plate. It’s not uncommon for microplate instruments to export data in this format.
Cq
1
2
3
4
5
6
A
24.180858612060547
20.74011993408203
17.183801651000977
13.774299621582031
10.29498291015625
6.967061996459961
B
24.15711784362793
20.77970314025879
17.171794891357422
13.768831253051758
10.362966537475586
6.870273113250732
C
24.238229751586914
20.78700828552246
17.147598266601563
13.779314041137695
10.292966842651367
6.735703945159912
Below is the code to load this data into a tidy
pandas.DataFrame
with the following columns:row: A letter identifying a row on the microplate, e.g. A-H
col: A number identifying a column on the microplate, e.g. 1-12
Cq: The \(C_q\) value measured for the identified well.
>>> import pandas as pd >>> def load_cq(path): ... return (pd ... .read_csv(path) ... .rename(columns={'Cq': 'row'}) ... .melt( ... id_vars=['row'], ... var_name='col', ... value_name='Cq', ... ) ... ) >>> data = load_cq('std_curve.csv') >>> data row col Cq 0 A 1 24.180859 1 B 1 24.157118 2 C 1 24.238230 3 A 2 20.740120 4 B 2 20.779703 5 C 2 20.787008 6 A 3 17.183802 7 B 3 17.171795 8 C 3 17.147598 9 A 4 13.774300 10 B 4 13.768831 11 C 4 13.779314 12 A 5 10.294983 13 B 5 10.362967 14 C 5 10.292967 15 A 6 6.967062 16 B 6 6.870273 17 C 6 6.735704
Use
wellmap.load()
to associate the labels specified in the TOML file (e.g. the dilutions and replicates) with the experimental data (e.g. the \(C_q\) values). This process has three steps:Load a data frame containing the data (see above).
Load another data frame containing the labels.
Merge the two data frames.
For the sake of clarity and completeness, we will first show how to perform these steps manually. Practically, though, it’s easier to let
wellmap
perform them automatically.Manual merge
Use the
wellmap.load()
function to create apandas.DataFrame
containing the information from the TOML file. This data frame will have columns for each label we specified: replicate, dilution. It will also have six columns identifying the wells in different ways: well, well0, row, col, row_i, col_j. These columns are redundant, but this redundancy makes it easier to merge the labels with the data. For example, if the wells are named “A1,A2,…” in the data, the well column can be used for the merge. If the wells are named “A01,A02,…”, the well0 column can be used instead. If the wells are named in some non-standard way, the row_i and col_j columns can be used to calculate an appropriate merge column.>>> import wellmap >>> labels = wellmap.load('std_curve.toml') >>> labels well well0 row col row_i col_j replicate dilution 0 A1 A01 A 1 0 0 1 100000.0 1 A2 A02 A 2 0 1 1 10000.0 2 A3 A03 A 3 0 2 1 1000.0 3 A4 A04 A 4 0 3 1 100.0 4 A5 A05 A 5 0 4 1 10.0 5 A6 A06 A 6 0 5 1 1.0 6 B1 B01 B 1 1 0 2 100000.0 7 B2 B02 B 2 1 1 2 10000.0 8 B3 B03 B 3 1 2 2 1000.0 9 B4 B04 B 4 1 3 2 100.0 10 B5 B05 B 5 1 4 2 10.0 11 B6 B06 B 6 1 5 2 1.0 12 C1 C01 C 1 2 0 3 100000.0 13 C2 C02 C 2 2 1 3 10000.0 14 C3 C03 C 3 2 2 3 1000.0 15 C4 C04 C 4 2 3 3 100.0 16 C5 C05 C 5 2 4 3 10.0 17 C6 C06 C 6 2 5 3 1.0
Use the
pandas.merge()
function to associate the labels with the data. In this case, both data frames have columns named row and col, sopandas
will automatically use those for the merge. It is also easy to merge using columns with different names; see the documentation onpandas.merge()
for more information.>>> import pandas as pd >>> df = pd.merge(labels, data) >>> df well well0 row col row_i col_j replicate dilution Cq 0 A1 A01 A 1 0 0 1 100000.0 24.180859 1 A2 A02 A 2 0 1 1 10000.0 20.740120 2 A3 A03 A 3 0 2 1 1000.0 17.183802 3 A4 A04 A 4 0 3 1 100.0 13.774300 4 A5 A05 A 5 0 4 1 10.0 10.294983 5 A6 A06 A 6 0 5 1 1.0 6.967062 6 B1 B01 B 1 1 0 2 100000.0 24.157118 7 B2 B02 B 2 1 1 2 10000.0 20.779703 8 B3 B03 B 3 1 2 2 1000.0 17.171795 9 B4 B04 B 4 1 3 2 100.0 13.768831 10 B5 B05 B 5 1 4 2 10.0 10.362967 11 B6 B06 B 6 1 5 2 1.0 6.870273 12 C1 C01 C 1 2 0 3 100000.0 24.238230 13 C2 C02 C 2 2 1 3 10000.0 20.787008 14 C3 C03 C 3 2 2 3 1000.0 17.147598 15 C4 C04 C 4 2 3 3 100.0 13.779314 16 C5 C05 C 5 2 4 3 10.0 10.292967 17 C6 C06 C 6 2 5 3 1.0 6.735704
Automatic merge
While it’s good to understand how the labels are merged with the data, it’s better to let
wellmap
perform the merge for you. Not only is this more succinct, it also handles some tricky corner cases behind the scenes, e.g. layouts with multiple data files.To load and merge the data using
wellmap.load()
, you need to provide the following arguments:data_loader: A function that accepts a path to a file and returns a
pandas.DataFrame
containing the data from that file. Note that the function we wrote in the previous section fulfills these requirements. If the raw data are tidy to begin with, it is often possible to directly usepandas.read_csv()
or similar for this argument.merge_cols: An indication of which columns to merge. In the snippet below,
True
means to use any columns that are shared between the two data frames (e.g. that have the same name). You can also use a dictionary to be more explicit about which columns to merge on.
Here we also provide the path_guess argument, which specifies that the experimental data can be found in a CSV file with the same base name as the layout. It also would’ve been possible to specify the path to the CSV directly from the TOML file (see meta.path), in which case this argument would’ve been unnecessary.
>>> df = wellmap.load( ... 'std_curve.toml', ... data_loader=load_cq, ... merge_cols=True, ... path_guess='{0.stem}.csv', ... ) >>> df well well0 row ... replicate dilution Cq 0 A1 A01 A ... 1 100000.0 24.180859 1 A2 A02 A ... 1 10000.0 20.740120 2 A3 A03 A ... 1 1000.0 17.183802 3 A4 A04 A ... 1 100.0 13.774300 4 A5 A05 A ... 1 10.0 10.294983 5 A6 A06 A ... 1 1.0 6.967062 6 B1 B01 B ... 2 100000.0 24.157118 7 B2 B02 B ... 2 10000.0 20.779703 8 B3 B03 B ... 2 1000.0 17.171795 9 B4 B04 B ... 2 100.0 13.768831 10 B5 B05 B ... 2 10.0 10.362967 11 B6 B06 B ... 2 1.0 6.870273 12 C1 C01 C ... 3 100000.0 24.238230 13 C2 C02 C ... 3 10000.0 20.787008 14 C3 C03 C ... 3 1000.0 17.147598 15 C4 C04 C ... 3 100.0 13.779314 16 C5 C05 C ... 3 10.0 10.292967 17 C6 C06 C ... 3 1.0 6.735704 [18 rows x 10 columns]
Analyze the data given the connection between the labels and the data. This step doesn’t involve
wellmap
, but is included here for completeness. The example below makes a linear regression of the data in log-space:#!/usr/bin/env python3 import wellmap import pandas as pd import numpy as np import matplotlib.pyplot as plt from scipy.stats import linregress def load_cq(path): return (pd .read_csv(path) .rename(columns={'Cq': 'row'}) .melt( id_vars=['row'], var_name='col', value_name='Cq', ) ) df = wellmap.load( 'std_curve.toml', data_loader=load_cq, merge_cols=True, path_guess='{0.stem}.csv', ) x = df['dilution'] y = df['Cq'] m, b, r, p, err = linregress(np.log10(x), y) x_fit = np.logspace(0, 5) y_fit = np.polyval((m, b), np.log10(x_fit)) r2 = r**2 eff = 100 * (10**(1/m) - 1) label = 'R²={:.5f}\neff={:.2f}%'.format(r2, eff) plt.plot(x_fit, y_fit, '--', label=label) plt.plot(x, y, '+') plt.legend(loc='best') plt.xscale('log') plt.xlabel('dilution') plt.ylabel('Cq') plt.show()