README.md

---
dataset_info:
  features:
  - name: pmid
    dtype: string
  - name: sentence
    dtype: string
  - name: cancer_type
    dtype: string
  - name: gene
    struct:
    - name: name
      dtype: string
    - name: pos
      sequence: int64
  - name: cancer
    struct:
    - name: name
      dtype: string
    - name: pos
      sequence: int64
  - name: CGE
    dtype: string
  - name: CCS
    dtype: string
  - name: PT
    dtype: string
  - name: IGE
    dtype: string
  - name: expression_change_keyword_1
    struct:
    - name: name
      dtype: string
    - name: pos
      sequence: int64
    - name: type
      dtype: string
  - name: expression_change_keyword_2
    struct:
    - name: name
      dtype: string
    - name: pos
      sequence: int64
    - name: type
      dtype: string
  splits:
  - name: train
    num_bytes: 361666
    num_examples: 821
  download_size: 99496
  dataset_size: 361666
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
license: cc-by-2.0
task_categories:
- text-classification
language:
- en
tags:
- biology
- cancer
- gene
- medical
pretty_name: CoMAGC
size_categories:
- n<1K
---
# Dataset Card for CoMAGC

## Dataset Description

- **Website:** http://biopathway.org/CoMAGC/
- **Paper:** [CoMAGC: a corpus with multi-faceted annotations of gene-cancer relations](https://bmcbioinformatics.biomedcentral.com/articles/10.1186/1471-2105-14-323)

#### Dataset Summary

<!-- Provide a quick summary of the dataset. -->
**CoMAGC Dataset Summary:**

CoMAGC is a corpus with multi-faceted annotations of gene-cancer relations.
CoMAGC consists of 821 sentences collected from MEDLINE abstracts, and the sentences are about three different types of cancers, or prostate, breast and ovarian cancers.
In CoMAGC, a piece of annotation is composed of four semantically orthogonal concepts that together express 1) how a gene changes, 2) how a cancer changes and 3) the causality between the gene and the cancer.
The four concepts that constitute the multi-faceted annotation scheme are Change in Gene Expression (CGE), Change in Cell State (CCS), Proposition Type (PT) and Initial Gene Expression level (IGE).

- CGE captures whether the expression level of a gene is `increased` or `decreased` in a cell
- CCS captures the way how the cell changes together with a gene expression level change
  - `normalTOnormal`: The cell or tissue remains as normal after the change in the gene’s expression level.
  - `normalTOcancer`: The cell or tissue acquires cancerous properties as the gene expression level changes; some cancerous properties are strengthened.
  - `cancerTOcancer`: There's no change in the cancerous properties of the cell or tissue despite the change in the expression level of the gene.
  - `cancerTOnormal`: The cell or tissue loses some cancerous properties as the gene expression level changes; some cancerous properties are weakened.
  - `unidentifiable`: The information about whether or not the gene expression level change accompanies cell or tissue state change is not provided. 
- PT captures whether the causality between the gene expression change and the cell property change
  - `observation`: Cell or tissue change accompanied by the gene expression level change is reported as observed but the causality between the two is not claimed. |
  - `causality`: The causality between the gene expression level change and the cell or tissue change is claimed.
- IGE captures the initial expression level of a gene before the change in its expression level
  - `up-regulated`: The initial gene expression level is higher than the expression level of the gene in the normal state.
  - `down-regulated`: The initial gene expression level is lower than the expression level of the gene in the normal state.
  - `unchanged`: The initial gene expression level is comparable to the expression level of the gene in the normal state.
  - `unidentifiable`: The information about the initial gene expression level is not provided.                                           |

The original dataset in XML format is available here: http://biopathway.org/CoMAGC/

We converted the dataset to a JSONL format before pushing the data to the hub.

### Languages
The language in the dataset is English.
## Dataset Structure
<!-- This section provides a description of the dataset fields, and additional information about the dataset structure such as criteria used to create the splits, relationships between data points, etc. -->
### Dataset Instances
An example of 'train' looks as follows:
```json
{
  "pmid": "11722842.s0",
  "sentence": "Isolation and characterization of the major form of human MUC18 cDNA gene and correlation of MUC18 over-expression in prostate cancer cell lines and tissues with malignant progression.",
  "cancer_type": "prostate",
  "gene": {
    "name": "MUC18",
    "pos": [93, 97]
  },
  "cancer": {
    "name": "prostate cancer",
    "pos": [118, 132]
  },
  "CGE": "increased",
  "CCS": "normalTOcancer",
  "PT": "observation",
  "IGE": "unchanged",
  "expression_change_keyword_1": {
    "name": "over-expression",
    "pos": [99, 113],
    "type": "Gene_expression"
  },
  "expression_change_keyword_2": {
    "name": "over-expression",
    "pos": [99, 113],
    "type": "Positive_regulation"
  }
}
```
### Data Fields
- `pmid`: the id of this sentence, a `string` feature.
- `sentence`: the text of this sentence, a `string` feature.
- `cancer_type`: the type of cancer in this sentence, a `string` feature.
- `gene`: gene entity
    - `pos`: character offsets of the gene entity, a list of `int32` features.
    - `name`: gene entity text, a `string` feature.
- `cancer`: cancer entity
    - `pos`: character offsets of the cancer entity, a list of `int32` features.
    - `name`: cancer entity text, a `string` feature.
- `CGE`: change in gene expression, a `string` feature.
- `CCS`: change in cell state, a `string` feature.
- `PT`: proposition type, a `string` feature.
- `IGE`: initial gene expression, a `string` feature.
- `expression_change_keyword_1`: a `dict`
  - `name`: keyword text, a `string` feature.
  - `pos`: character offsets of the keyword, a list of `int32` features.
  - `type`: type of the expression change keyword, a `string` feature.
- `expression_change_keyword_2`: a `dict`
  - `name`: keyword text, a `string` feature.
  - `pos`: character offsets of the keyword, a list of `int32` features.
  - `type`: type of the expression change keyword, a `string` feature.

## Citation
<!-- If there is a paper or blog post introducing the dataset, the APA and Bibtex information for that should go in this section. -->
**BibTeX:**
```
@article{lee2013comagc,
  title={CoMAGC: a corpus with multi-faceted annotations of gene-cancer relations},
  author={Lee, Hee-Jin and Shim, Sang-Hyung and Song, Mi-Ryoung and Lee, Hyunju and Park, Jong C},
  journal={BMC bioinformatics},
  volume={14},
  pages={1--17},
  year={2013},
  publisher={Springer}
}
```
**APA:**
- Lee, H. J., Shim, S. H., Song, M. R., Lee, H., & Park, J. C. (2013). CoMAGC: a corpus with multi-faceted annotations of gene-cancer relations. BMC bioinformatics, 14, 1-17.
## Dataset Card Authors
[@phucdev](https://github.com/phucdev)