![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2224301%2Fd1665c030bca1d7b4741a4f6e5de25e7%2Frna.png?generation=1742588393597198&alt=media\")
Let's understand the data
\nWe are provided with this dataset
\n[train/validation/test]_sequences.csv - the target sequences of the RNA molecules.
target_id: is the index at protein databank with the chain ID
sequence: string representation of the nucleotides
\ntemporal_cutoff: date in which the sequence was published. It is advised by the organizer to train only before 2022-05-27 to have an additional validation data.
description: Details of the origin of the sequence
\nall_sequences: This has the label of the sequence, the structure how it is fold and where it came from. It also contains the actual string of nucleotides.
Let's look at an example:
\n SCL_A\n GGGUGCUCAGUACGAGAGGAACCGCACCC\n --\n THE SARCIN-RICIN LOOP, A MODULAR RNA\n >SCL_1|Chain A|RNA SARCIN-RICIN LOOP|Rattus norvegicus ()\\\\nGGGUGCUCAGUACGAGAGGAACCGCACCC\\\\n'\n3D structure
\n
Chain and atomic clashes
\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2224301%2F13a23a6fdf2f944ab1d6ea7c7e67d4ae%2Fchain_rna.png?generation=1742588409703656&alt=media\")
[train/validation]_labels.csv
\nThis contains the coordinates of the 3D strructure of the protein in each of the nuclotides, our training examples only have one position in space x_1, y_1, z_1. In reality, these are dynamic, meaning that their position changes, and methods can capture more than one conformation. These are represented in the validation data, the submission file can only have five different conformation.
Goal:
\nOur task is based on the chain predict the position of each nucleotide in a 3dimensional space, in other words, how it will fold.
\nDomain ideas:
\nLet's understand the data
\nWe are provided with this dataset
\n[train/validation/test]_sequences.csv - the target sequences of the RNA molecules.
target_id: is the index at protein databank with the chain ID
sequence: string representation of the nucleotides
\ntemporal_cutoff: date in which the sequence was published. It is advised by the organizer to train only before 2022-05-27 to have an additional validation data.
description: Details of the origin of the sequence
\nall_sequences: This has the label of the sequence, the structure how it is fold and where it came from. It also contains the actual string of nucleotides.
Let's look at an example:
\n SCL_A\n GGGUGCUCAGUACGAGAGGAACCGCACCC\n --\n THE SARCIN-RICIN LOOP, A MODULAR RNA\n >SCL_1|Chain A|RNA SARCIN-RICIN LOOP|Rattus norvegicus ()\\\\nGGGUGCUCAGUACGAGAGGAACCGCACCC\\\\n'\n3D structure
\n
Chain and atomic clashes
\n
[train/validation]_labels.csv
\nThis contains the coordinates of the 3D strructure of the protein in each of the nuclotides, our training examples only have one position in space x_1, y_1, z_1. In reality, these are dynamic, meaning that their position changes, and methods can capture more than one conformation. These are represented in the validation data, the submission file can only have five different conformation.
Goal:
\nOur task is based on the chain predict the position of each nucleotide in a 3dimensional space, in other words, how it will fold.
\nDomain ideas:
\nThe dataset consists of RNA sequence data with the following key components:
\ntarget_id: Links the RNA sequence to a protein structure in the Protein Data Bank (PDB) by its chain ID.
\nsequence: The nucleotide sequence of the RNA molecule.
\ntemporal_cutoff: The publication date of the sequence, with a suggestion to train only on sequences published before 2022-05-27 for additional validation.
\ndescription: Provides context on the sequence's origin.
\nall_sequences: Includes the sequence label, structural information, and the nucleotide string.
\nThis structure will help in analyzing RNA sequences and their relationships to protein structures.
", "rawMarkdown": "The dataset consists of RNA sequence data with the following key components:\n\ntarget_id: Links the RNA sequence to a protein structure in the Protein Data Bank (PDB) by its chain ID.\n\nsequence: The nucleotide sequence of the RNA molecule.\n\ntemporal_cutoff: The publication date of the sequence, with a suggestion to train only on sequences published before 2022-05-27 for additional validation.\n\ndescription: Provides context on the sequence's origin.\n\nall_sequences: Includes the sequence label, structural information, and the nucleotide string.\n\nThis structure will help in analyzing RNA sequences and their relationships to protein structures.", "votes": null } ], "comments": [ { "id": 3158943, "author_name": "atharvasoundankar", "author_url": "", "post_date": "03/25/2025 04:50:16", "content": "The dataset consists of RNA sequence data with the following key components:
\ntarget_id: Links the RNA sequence to a protein structure in the Protein Data Bank (PDB) by its chain ID.
\nsequence: The nucleotide sequence of the RNA molecule.
\ntemporal_cutoff: The publication date of the sequence, with a suggestion to train only on sequences published before 2022-05-27 for additional validation.
\ndescription: Provides context on the sequence's origin.
\nall_sequences: Includes the sequence label, structural information, and the nucleotide string.
\nThis structure will help in analyzing RNA sequences and their relationships to protein structures.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "3156167": "Let's understand the data\nWe are provided with this dataset\n**[train/validation/test]_sequences.csv** - the target sequences of the RNA molecules.\n\n`target_id`: is the index at protein databank with the chain ID\n\n`sequence`: string representation of the nucleotides\ntemporal_cutoff: date in which the sequence was published. It is advised by the organizer to train only before 2022-05-27 to have an additional validation data.\n\n`description`: Details of the origin of the sequence\n`all_sequences`: This has the label of the sequence, the structure how it is fold and where it came from. It also contains the actual string of nucleotides.\n\nLet's look at an example:\n\n```\ntarget_id 1SCL_A\nsequence GGGUGCUCAGUACGAGAGGAACCGCACCC\ntemporal_cutoff 1995-01-26\ndescription THE SARCIN-RICIN LOOP, A MODULAR RNA\nall_sequences >1SCL_1|Chain A|RNA SARCIN-RICIN LOOP|Rattus norvegicus (10116)\\\\nGGGUGCUCAGUACGAGAGGAACCGCACCC\\\\n'\n\n```\n\n- Protein databank identifier: is 1SCL_A https://www.rcsb.org/structure/1SCL\n- Nucleotide's sequence: GGGUGCUCAGUACGAGAGGAACCGCACCC\n- It was discover in 1995-01-26\n- It is a Chain A which means it is part of a bigger molecular structure\n- **RNA SARCIN-RICIN LOOP:** This is the descriptive name of the RNA molecule, indicating its structure and function.\n- **Rattus norvegicus (10116):** This specifies the organism from which the RNA was obtained (in this case, a rat) and its taxonomy ID.\n\n3D structure\n\n\n\nChain and atomic clashes\n\n\n\n`[train/validation]_labels.csv`\nThis contains the coordinates of the 3D strructure of the protein in each of the nuclotides, our training examples only have one position in space x_1, y_1, z_1. In reality, these are dynamic, meaning that their position changes, and methods can capture more than one conformation. These are represented in the validation data, the submission file can only have five different conformation.\n\nGoal:\n\nOur task is based on the chain predict the position of each nucleotide in a 3dimensional space, in other words, how it will fold. \n\nDomain ideas:\n\n- \"Atomic clashes\" refer to situations where atoms are too close to each other in a molecule or a protein structure, violating the rules of how atoms should normally occupy space. This can signal modeling errors or high structural stress.", "3158943": "The dataset consists of RNA sequence data with the following key components:\n\ntarget_id: Links the RNA sequence to a protein structure in the Protein Data Bank (PDB) by its chain ID.\n\nsequence: The nucleotide sequence of the RNA molecule.\n\ntemporal_cutoff: The publication date of the sequence, with a suggestion to train only on sequences published before 2022-05-27 for additional validation.\n\ndescription: Provides context on the sequence's origin.\n\nall_sequences: Includes the sequence label, structural information, and the nucleotide string.\n\nThis structure will help in analyzing RNA sequences and their relationships to protein structures." }, "source": "meta" }