Help

Do you produce scientific findings using programming languages such as Python or R? Then you may be interested to learn about the TIB Knowledge Loom - an emerging Open Science Digital Library that can increase the impact of your research. By making a few tweaks to your existing code, you can ensure your scientific findings are machine readable and reproducible not only by humans, but also by machines. This means that with a click of a button in the TIB Knowledge Loom, researchers can download your results as data ready for immediate reuse.

The recipe is straightforward:

Step 1. Install the dtreg package for Python or R;

Step 2. Describe your scientific findings in a machine-reusable form using the dtreg package. The table below guides the selection of data types. If none of the data types meet your needs or you require troubleshooting assistance contact us at knowledgeloom@tib.eu;

Step 3. Submit your loom article data together with the manuscript and supplementary research code and data and we will take care of the curation and the publication process. If you don't want to create the loom article data yourself, just Submit your article, data, and code - we will do it for you.

Let's see how it's done.

Example

You are researching Irises and want to know whether the setosa and virginica species have different petal length.

Step 1. Install the dtreg package you need.

Python R

pip install dtreg

install.packages("dtreg")

Next, you conduct the t-test using your preferred scripting language.

Python R

from sklearn import datasets
from scipy.stats import ttest_ind
import pandas as pd

iris = datasets.load_iris(as_frame=True).frame
iris['Species'] = iris['target'].replace({0: 'setosa', 1: 'versicolor', 2: 'virginica'})
setosa = iris[iris['Species'] == 'setosa']['petal length (cm)']
virginica = iris[iris['Species'] == 'virginica']['petal length (cm)']
tt = ttest_ind(setosa, virginica, equal_var=False)
df_results = pd.DataFrame({"t.statistic": [tt.statistic],
                           "df": [tt.df],
                           "p.value": [tt.pvalue]})

library(dplyr)

data(iris)
setosa <- iris |>
  dplyr::filter(Species == "setosa") |>
  dplyr::select(Petal.Length)
virginica <- iris |>
  dplyr::filter(Species == "virginica") |>
  dplyr::select(Petal.Length)

tt <- stats::t.test(setosa, virginica, var.equal = FALSE)
df_results = data.frame(
  t.statistic = tt$statistic,
  df = tt$parameter,
  p.value = tt$p.value
)
rownames(df_results) <- "value"

Step 2. In your manuscript, you might include a boxplot of the data and state that the p-value is less than 0.001. Including this kind of plot in your published article only presents the data in a form that is reusable to humans (i.e., a computer can not interpret and organize this information). But, with a few small additions to your code, you can express your findings in machine-readable form as well. To do so, you use the dtreg package and the data type for group comparison (https://doi.org/21.T11969/b9335ce2c99ed87735a6) in your script.

Python R


from dtreg.load_datatype import load_datatype
from dtreg.to_jsonld import to_jsonld

dt_da = load_datatype("https://doi.org/21.T11969/feeb33ad3e4440682a4d") # Data Analysis
dt_gc = load_datatype("https://doi.org/21.T11969/b9335ce2c99ed87735a6") # Group Comparison

inst_da = dt_da.data_analysis(
  has_part=dt_gc.group_comparison(
    label="t-test Iris petal length setosa vs virginica",
    executes=dt_gc.software_method(
      label="ttest_ind",
      is_implemented_by="ttest_ind(setosa, virginica, equal_var = False)",
      part_of=dt_gc.software_library(
        label="scipy",
        version_info="1.15.1",
        part_of=dt_gc.software(label="Python",
                               version_info="3.12.5")
      )
    ),
    targets=dt_gc.component(label="petal length (cm)"),
    has_input=dt_gc.data_item(label="Iris petal length setosa virginica",
                              source_table=iris),
    has_output=dt_gc.data_item(source_table=df_results)
  )
)

json = to_jsonld(inst_da)

with open("data-analysis-1.json", "w") as f:
  f.write(json)

library(dtreg)

dt_da <- dtreg::load_datatype("https://doi.org/21.T11969/feeb33ad3e4440682a4d") # Data Analysis
dt_gc <- dtreg::load_datatype("https://doi.org/21.T11969/b9335ce2c99ed87735a6") # Group Comparison

inst_da <- dt_da$data_analysis(
  has_part = dt_gc$group_comparison(
    label = "t-test Iris petal length setosa vs virginica",
    executes = dt_gc$software_method(
      label = "t.test",
      is_implemented_by = "stats::t.test(setosa, virginica, var.equal = FALSE)",
      part_of = dt_gc$software_library(
        label = "stats",
        version_info = "4.3.1",
        part_of = dt_gc$software(label = "R",
                                 version_info = "4.3.1")
      )
    ),
    targets = dt_gc$component(label = "Petal.Length"),
    has_input = dt_gc$data_item(label = "Iris petal length setosa virginica",
                                source_table = iris),
    has_output = dt_gc$data_item(source_table = df_results)
  )
)

json <- dtreg::to_jsonld(inst_da)

write(json, "data-analysis-1.json")

Step 3. The machine-readable expression of your t-test is now in the file named data-analysis-1.json. When you submit your manuscript to your favourite journal or conference get in touch with us at knowledgeloom@tib.eu and we will support the publication of this supplementary data.

Data types

Nr	Data type	Definition	Methods (examples)
1	https://doi.org/21.T11969/feeb33ad3e4440682a4d (Data Analysis)	Data transformation aimed at analyzing a dataset	Composition of methods
2	https://doi.org/21.T11969/37182ecfb4474942e255 (Data Preprocessing)	Data transformation aimed at preparing a dataset for further analysis	Filtering, missing data imputation
3	https://doi.org/21.T11969/5b66cb584b974b186f37 (Descriptive Statistics)	Data transformation aimed at describing characteristics of a dataset, such as the central tendency or variability	Mean, median, standard deviation
4	https://doi.org/21.T11969/5e782e67e70d0b2a022a (Algorithm Evaluation)	Data transformation aimed at evaluating the performance of an algorithm for a specified task on a dataset	Benchmark-based model evaluation (Task, Dataset, Metric, Score)
5	https://doi.org/21.T11969/c6b413ba96ba477b5dca (Multilevel Analysis)	Data transformation on multilevel data with models including fixed and random effects	Hierarchical linear model / linear mixed model, multilevel structural equation model, multilevel latent class analysis
6	https://doi.org/21.T11969/3f64a93eef69d721518f (Correlation Analysis)	Data transformation aimed at evaluating the strength and direction of the relationship between variables	Pearson correlation, Spearman correlation
7	https://doi.org/21.T11969/b9335ce2c99ed87735a6 (Group Comparison)	Data transformation aimed at comparing means of two or more groups	t-test, ANOVA, MANOVA, ANCOVA, Chi-square, Wilcoxon, Kruskal Wallis, Least Square Means
8	https://doi.org/21.T11969/286991b26f02d58ee490 (Regression Analysis)	Data transformation aimed at exploring the relationship between dependent/target variable(s) and independent/input variable(s)	Simple / multiple linear regression, regression in machine learning
9	https://doi.org/21.T11969/6e3e29ce3ba5a0b9abfe (Class Prediction)	Data transformation aimed at predicting a categorical/ordinal class label given the input data	Logistic regression, ordinal regression, classification in machine learning
10	https://doi.org/21.T11969/c6e19df3b52ab8d855a9 (Class Discovery)	Data transformation aimed at discovering classes or clusters in unlabeled data	Clustering, latent class analysis
11	https://doi.org/21.T11969/437807f8d1a81b5138a3 (Factor Analysis)	Data transformation aimed at discerning latent factors in the data	Exploratory factor analysis, confirmatory factor analysis, structural equation model

Additional resources

Vignette for the dtreg R package with additional examples