metbit

Metbit is a Python package designed for the analysis of metabolomics data. It provides a range of tools and functions to process, visualize, and interpret metabolomics datasets. With Metbit, you can perform various statistical analyses, identify biomarkers, and generate informative visualizations to gain insights into your metabolomics experiments. Whether you are a researcher, scientist, or data analyst working in the field of metabolomics, Metbit can help streamline your data analysis workflow and facilitate the interpretation of complex metabolomics data.

How to install

pip install metbit

Features:

1. PCA model and visualisation

2. OPLS-DA model and visualisation

3. STOCSY and STOCSY app

Example:

Principal component analysis

PCA is used to transform a large set of variables into a smaller one that still contains most of the information in the large set. This is particularly useful when dealing with high-dimensional data, where visualizing and analyzing the data can be challenging.

To perform Principal Component Analysis (PCA) in a Python environment, you can use the metbit library. Here's a step-by-step guide to import the necessary package and perform PCA:

import pandas as pd
from metbit import pca

df = pd.read_csv("metbit_tutorial_data.csv")

Use .iloc or df.head(10)[df.columns[:10]] to display 10 rows and 10 columns first

df.iloc[:10, :10]

Output:

Group	Time point	0.0	0.0001716414	0.0003432828	0.0005149242	0.0006865656	0.000858207	0.0010298484	0.0012014898
Group B	3	3024.2	3923.9	4758.23	4551.28	3737.53	3469.81	3646.49	3278.41
Group A	3	3776.08	3441.18	3479.89	4102.29	5089.12	6000.92	6556.49	6687.83
Group A	2	3823.99	3227.06	2793.23	2544.2	2254.06	1843.89	1470.21	1362.43
Group B	2	21926	21546.2	21155.6	20190.6	18755.6	17993.4	18545.5	19496.6
Group A	4	2997.22	2130.68	1993.8	2948.87	4414.49	5267.69	4897.94	3868.98
Group B	4	12988.5	14361.2	15288.7	15439.6	15410.4	15513.2	15528	15446.5
Group A	2	202.293	111.968	640.931	1732.62	2926.78	3299.18	2308.54	783.053
Group A	3	4813.91	4822.44	4153.81	2861.74	1405.9	575.416	725.433	1315.78
Group A	4	34822.5	34380.2	33536.9	32369.3	31296.5	30737.1	30694.3	30909.4
Group A	4	124.841	677.809	841.232	659.092	479.715	438.279	323.827	-43.9303

Assign object to perform PCA

X: data frame of features to test

features_name: features name of X data frame

color_: series of group to label with color

symbol_: series of time point to label with symbol

time_order: assign order of symbol

X = df.iloc[:, 2:]
features_name = X.columns.astype(float).to_list()
color_ = df["Group"]
symbol_ = df["Time point"]
time_order = {1:0, 2:1, 3:2, 4:2}

Assign and fit PCA model

pca_mod = pca(X=X, label=color_, features_name=ppm, n_components=3)
pca_mod.fit()

Visualisation

pca_mod.plot_cumulative_observed()

Output:

pca_mod.plot_pca_scores(pc=["PC1", "PC2"], symbol_=symbol_)

Output:

pca_mod.plot_pca_scores(pc=["PC1", "PC3"], symbol_=symbol_)

Output:

pca_mod.plot_3d_pca(marker_size=10, symbol_=symbol_)

Output:

To observe time series of PCA you can perform times trajectory plot use function plot_trajectory

pca_mod.plot_pca_trajectory(time_=symbol_, time_order=time_order, pc=["PC1", "PC2"])

Output:

pca_mod.plot_pca_trajectory(time_=symbol_, time_order=time_order, pc=["PC1", "PC3"])

Output:

Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA)

Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA) was proposed by Prof. Svante Wold in 2002 as a variant of PLS-DA, using a mathematical filter to remove systematic variance unrelated to the sample class. This is particularly advantageous in metabolomics, such as distinguishing the metabolomic signature of coronary disease without confounding factors like sex. However, OPLS-DA is less common than PLS-DA due to increased risk of overfitting and its limitation to binary classification.

from metbit import opls_da 
import pandas as pd 

1. Load the data and data manipulation

df = pd.read_csv("metbit_tutorial_data.csv")
#Exclude base line (Time point 1)
df.drop(df.loc[df["Time point"]==1].index, inplace=True)

X = df.iloc[:, 2:]
ppm = X.columns.astype(float).to_list()
y = df["Group"]

opls_da_mod = opls_da(X=X, y=y, features_name=ppm, scale='uv', auto_ncomp=True)

opls_da_mod.fit()

Output:

OPLS-DA model is fitted in 2.5721652507782 seconds

Visualisation

opls_da_mod.plot_oplsda_scores()

Output:

opls_da_mod.plot_loading()

Output:

opls_da_mod.plot_s_scores()

Output:

opls_da_mod.permutation_test(n_permutations=100, n_jobs=-1)

Output:

[Parallel(n_jobs=-1)]: Using backend LokyBackend with 8 concurrent workers.
[Parallel(n_jobs=-1)]: Done   2 tasks      | elapsed:    8.5s
[Parallel(n_jobs=-1)]: Done   9 tasks      | elapsed:   11.1s
[Parallel(n_jobs=-1)]: Done  16 tasks      | elapsed:   13.2s
[Parallel(n_jobs=-1)]: Done  25 tasks      | elapsed:   17.5s
[Parallel(n_jobs=-1)]: Done  34 tasks      | elapsed:   20.8s
[Parallel(n_jobs=-1)]: Done  45 tasks      | elapsed:   23.7s
[Parallel(n_jobs=-1)]: Done  56 tasks      | elapsed:   27.4s
[Parallel(n_jobs=-1)]: Done  69 tasks      | elapsed:   33.5s
[Parallel(n_jobs=-1)]: Done  82 tasks      | elapsed:   37.9s
[Parallel(n_jobs=-1)]: Done  96 out of 100 | elapsed:   42.7s remaining:    1.8s


Permutation test is performed in 46.19982290267944 seconds

[Parallel(n_jobs=-1)]: Done 100 out of 100 | elapsed:   43.6s finished

opls_da_mod.plot_hist()

Output:
![opls da permutation histogram](./src/img/oplsda_hist.svg)

``` python
opls_da_mod.vip_scores()
opls_da_mod.vip_plot(threshold=2)

Output:

---

Publication:

Karunasumetta C, Tourthong W, Mala R, Chatgasem C, Bubpamala T, Punchai S, Sawanyawisuth K. Comparative Analysis of Metabolomic Responses in On-Pump and Off-Pump Coronary Artery Bypass Grafting. Ann Thorac Cardiovasc Surg. 2024;30(1):24-00126. doi: 10.5761/atcs.oa.24-00126. PMID: 39631940; PMCID: PMC11634389.