DV2S: DNA Variant to Structure

A computational tool that maps DNA sequence variations to protein structures, enabling structural interpretation of genetic variants.

Features

• DNA to Protein Mapping: Translate DNA sequences and map variants to protein structures
• Multiple Structure Input: Support for PDB and mmCIF formats
• Flexible Operation Modes:
  • consensus: Generate consensus sequence for structure prediction
  • map: Map alignment to existing protein structure
  • skip: Skip structure processing
• Advanced Structure Prediction: Integration with ESM, Boltz-2, and AlphaFold2
• Quality Control: pLDDT score filtering for predicted structures

Installation

# install for all user
pip install dv2s
# install for current user
pip install dv2s --user

Quick Start

python3 -m dv2s -dna sequences.fasta -nvidia_key key_file -output result

Usage

# use consensus sequence of alignment to predict protein structure and analyze
python3 -m dv2s -dna dna_seq.fasta -output analysis_results
# use nvidia's API to predict protein structure via Boltz-2 model
python3 -m dv2s -dna dna_seq.fasta -output analysis_results -nvidia_key key_file -predict boltz-2
# map DNA alignment to given protein structure's sequence
python3 dv2s.py -dna sequences.fasta -pdb structure.pdb -mode map
# only preprocess and align the input sequence, skip protein strucutre prediction and analyze
python3 dv2s.py -dna sequences.fasta -mode skip
# use previous protein alignment to skip the align step
python3 dv2s.py -dna sequences.fasta -protein_aln sequence.protein.aln -nvidia_key key_file -predict esm-long

Command Line Options

Sequence Input

• -dna Required: DNA sequences in FASTA format
• -protein_aln: Aligned protein sequences in FASTA format
• -table_id: Translation table ID (default: 1, standard genetic code)

Structure Input

• -mode: Operation mode (consensus, map, skip), default: consensus
  • consensus: use alignment to generate consensus sequence and generate structure prediction
  • map: map alignment to given protein structure
  • skip: skip structure prediction and analysis
• -pdb: Protein structure in PDB format
• -mmcif: Protein structure in mmCIF format
• -predict: Structure prediction method (auto, esm, esm-long, boltz-2, alphafold2)
  • auto: Try all methods
  • esm: Use ESMFold server, only suitable for protein sequence shorter than 400 aa
  • esm-long: Use nvidia's ESMFold API, allow longer input (shorter than 1024 aa)
  • boltz-2: Use nvidia's Boltz-2 API, allow the longest input (shorter than 4096 aa)
  • alphafold2: Use nvidia's AlphaFold2 API, allow the longest input (shorter than 4096 aa) but slower
• -nvidia_key: nvidia API key file for prediction. Text file that contains only one line for the API key

Options

• -mask_low_plddt: Mask low pLDDT score residues in predicted structures. Set the residues' B-factor to 0
• -min_plddt: Minimum pLDDT value threshold (default: 0.3)
• -n_thread: Number of threads (default: -1, use all CPU cores)
• -output: Output directory for results
• -gene: Gene name, for retrieve protein structure from Uniprot
• -organism: Organism name (e.g., "Oryza sativa"), for Uniprot. Currently, Uniprot may return unwanted result.

Output

DV2S use input filename's prefix as output's prefix. DV2S generates comprehensive outputs including:

Summary

• .csv: CSV-format result of the analysis

Sequence

• .dna.fasta: A copy of input DNA sequences
• .protein.fasta: Translated protein sequences
• .dna_cons.fasta: Consensus sequence generated from the DNA alignment without gaps
• .protein_cons.fasta: Consensus sequence generated from the protein alignment without gaps

Alignment

• .dna.aln: DNA alignment, generated from input DNA sequences and protein alignment
• .protein.aln: Protein alignment
• .clean_dna.aln: DNA alignment that exclude invalid protein-coding sequences
• .clean_protein.aln: Protein alignment that exclude invalid protein-coding sequences

Subalign

• .helix_DNA.aln: A subset of sites in DNA alignment which correspond to protein helix structure. The secondary structure is defined by the consensus sequence's protein structure
• .strand_DNA.aln: A subset of sites in DNA alignment which correspond to protein strand structure
• .coil_DNA.aln: A subset of sites in DNA alignment which correspond to protein coil structure
• .helix_protein.aln: A subset of sites in protein alignment which correspond to protein helix structure
• .strand_protein.aln: A subset of sites in protein alignment which correspond to protein strand structure
• .coil_protein.aln: A subset of sites in protein alignment which correspond to protein coil structure

Protein structure

• .predict.pdb: Protein structure prediction result
• .Consensus_ratio.mmcif: MMCIF-format protein structure file, the B-factor column contains normalized consensus ratio of the DNA alignment
• .DNA_entropy.mmcif: MMCIF-format protein structure file, the B-factor column contains normalized DNA entropy of the DNA alignment
• .DNA_Pi.mmcif: MMCIF-format protein structure file, the B-factor column contains normalized nucleotide diversity (Pi) of the DNA alignment
• .DNA_Pi_omega.mmcif: MMCIF-format protein structure file, the B-factor column contains normalized Pi-omega (nonsynonymously variance rate/synonymous variance rate) of the DNA alignment. Inf and NaN are normalized to 0 or 5 * max value for visualization
• .protein_entropy.mmcif: MMCIF-format protein structure file, the B-factor column contains normalized protein entropy of the DNA alignment
• .protein_Pi.mmcif: MMCIF-format protein structure file, the B-factor column contains normalized nucleotide diversity (Pi) of the protein alignment

Others

• .log: DV2S log file
• .dssp.log: DSSP log file
• .mafft.log: MAFFT log file

Dependencies

• Python 3.12+
• Structure prediction tools or APIs (ESM, AlphaFold2, etc.)

Citation

If you use DV2S in your research, please cite:

[Citation information to be added]

License

This project is licensed under the APL-3 License - see the LICENSE file for details.

Support

For questions and support, please open an issue on GitHub.