Readme

What is a Structural Alphabet?

A structural alphabet is a way to represent protein structures using a discrete set of symbols, similar to how amino acids represent protein sequences. These symbols capture local backbone conformations, allowing complex 3D structures to be encoded as linear sequences. This alphabet is generated by Foldseek.

How Can These Characters Be Used to Build Alignments?

By converting 3D structures into sequences of structural characters, traditional sequence alignment methods can be applied to compare protein structures. This enables multiple structure alignments written out as MSAs and subsequent downstream analysis.

Why Are Correct Alignments Important?

Alignments determine how structures are compared and analyzed. Poor alignments can lead to incorrect evolutionary interpretations, misidentification of structural similarities, and errors in downstream applications like phylogenetic analysis.

What Does This App Do?

Structome-AlignViewer allows users to visualize 3Di character alignments alongside molecular structures. It provides tools to inspect alignment quality using structure-based visualization to ensure accuracy in structure-based evolutionary analyses.

Input Data Methods

The application now supports four different input methods, available via tabs on the input page:

• By PDB ID: Provide a semicolon-delimited list of PDB chain IDs (e.g., 1hv4_A;1hv4_B). The server will download the corresponding structures from the RCSB PDB.
• By Amino Acid Sequence: Provide two or more protein sequences in FASTA format. The server will use the ESMFold API to predict the 3D structure for each sequence before alignment.
• By Custom Structure Dataset: Upload a .zip archive containing multiple single-chain structure files in mmCIF (.cif) format. This allows for the analysis of custom structures.
• By Pre-existing Alignment: Paste a complete, pre-computed alignment in Clustal format but could be sourced from anywhere. This bypasses the structure acquisition and alignment steps and proceeds directly to analysis. Please note that this app has no way of testing if the alignment is of amino acids or 3Di characters. This app is designed for 3Di character alignments only and will assume any alignment to comprise 3Di characters only.

For methods that generate a new alignment, users can choose between the ClustalW and MAFFT alignment algorithms. As of 1st Sept, 2025 Z-scores profiles are being constructed for SCOP and CATH alignmnets. Once they are ready, MAFFT will be able to assess alignment scores based on these backgrounds.

What to Look for in Structures Before Providing Them to Structome-AlignViewer

• Ensure PDB chain accessions are valid/current.
• While attempts have been made to account for various types of errors, ensure structures are complete will help avoid unexpected errors.

Results Page

The results page displays:

• A protein structure visualizer. For jobs that use 3D structures, a viewer is provided with a dropdown menu to toggle between them.
• An interactive feature viewer that displays the 3Di alignment. Clicking a column in the alignment will highlight the corresponding residue in the 3D structure viewer. Columns with low coverage (<50% non-gap characters) that are removed in the trimmed alignment are marked with a red bar.
• A Job Information panel displaying metadata about the submission, including the input type and alignment method used.
• A Column Statistics bar chart visualizing the total number of columns in the alignment versus the number of valid and dropped columns.

Per-Column Confidence Score and Global Confidence Score

When analyzing alignments, it is crucial to assess the reliability of each column. The Confidence score is a proxy for identifying regions where structures are well-aligned versus areas of uncertainty. This score is derived from the substitution matrix, which assigns values to character replacements based on their likelihood in structural alignments.Higher scores indicate greater agreement between structures, while lower scores suggest high variability or misalignment. The Average (Avg) provides an overall measure of alignment quality by averaging the per-column confidence scores across only the well-aligned columns (after filtering out low-information regions). This ensures that the global score reflects the alignment's reliability rather than being biased by poor regions. A higher global confidence score suggests a well-conserved alignment, while a lower score may indicate inconsistent structural relationships. This information can be used to refine alignments or highlight regions needing further inspection.

Update 26th March, 2025 : The confidence per column is now encoded as b-factor and visualised on each structure. These structures can now be downloaded. Low confidence regions appear as blue (score approaching 0), whereas high confidence regions appear red (score approaching 1).

Alignment Statistics Panel

The statistics panel summarizes the quality of the alignment. This panel provides:

• Assessment: A qualitative label (e.g., Very High Confidence) derived from average Z-score interpretation (from SCOP and CATH).
• Average Confidence: Mean of all normalized per-column scores based on substitution matrix comparisons.
• Z-score (SCOP): Confidence relative to SCOP (family-level) alignment distributions.
• Z-score (CATH): Confidence relative to CATH (homology-level) alignment distributions.

These statistics help users quickly assess the reliability and structural consistency of an alignment.

What Can Be Downloaded?

Users can download:

• The 3Di sequences in a single file.
• The full 3Di alignment excluding characters reserved for amino acid alignments (*, :, .).
• A trimmed alignment. Similar to Gblock output, the trimmed alignment only has columns with >50% non-gap characters.
• The substitution matrix used to generate the alignment and supporting data files used in calculations
• Update 26th March, 2025 : The structures with their b-factor column modified can now be downloaded.
• Update 1st Sept, 2025 : Alignments can now be constructed using MAFFT, in addition to CLUSTAL and these can also be downloaded.

What is planned next for this resource?

One modifications is planned.

• Ability to use a custom substitution matrix

This enhancements will be rolled out soon-ish. 🤞

Citation

If you use Structome-AlignViewer in your work, please cite:

@article{malik2025structome,
  title   = {Structome-AlignViewer: On Confidence Assessment in Structure-Aware Alignments},
  author  = {Malik, Ashar J and Mao, Siying and Hugenholtz, Philip J and Ascher, David B},
  journal = {bioRxiv},
  pages   = {2025--05},
  year    = {2025},
  publisher = {Cold Spring Harbor Laboratory}
}

(The peer-reviewed version will be linked here once available.)

Acknowledgement

Structome-AlignViewer would like to thank the anonymous reviewer for its most recent upgrade.

Contact

For any issues with the server or functionality, please email and include job IDs.

Ashar Malik

Email: ashar.malik@uq.edu.au