Do "Newly Born" orphan proteins resemble "Never Born" proteins? A study using three deep learning algorithms.

Area of Science:

• * Structural biology
• * Computational biology
• * Bioinformatics

Background:

• * Orphan proteins, lacking homology, arise from novel gene expression and appear across evolution.
• * Distinguishing between structured and intrinsically disordered proteins is crucial for understanding their function.
• * Recent advancements in protein structure prediction offer new tools for analyzing these unique proteins.

Purpose of the Study:

• * To evaluate the utility of AlphaFold2, RoseTTAFold, and ESMFold in predicting structures of orphan and "Never Born" proteins.
• * To compare predicted structures with experimental data for validation.
• * To assess the potential of these tools for characterizing novel protein folds and functions.

Main Methods:

• * Utilized AlphaFold2, RoseTTAFold, and ESMFold to predict structures of "Never Born" protein groups (Group 1: structured, Group 3: intrinsically disordered).
• * Compared predicted structures of orphan proteins with known crystal structures.
• * Analyzed predicted structures for seven orphan proteins with unknown 3D structures.

Main Results:

• * All three algorithms accurately predicted compact structures for Group 1 ("Never Born") proteins and extended structures for Group 3 (intrinsically disordered) proteins, aligning with experimental data.
• * Predictions for a taxonomically restricted protein with a known fold closely matched its crystal structure.
• * For orphan proteins with unknown structures, predictions generally aligned with sequence-based disorder predictions, with most predicted as compact.

Conclusions:

• * AlphaFold2, RoseTTAFold, and ESMFold demonstrate capability in differentiating protein structural characteristics, including distinguishing between structured and intrinsically disordered proteins.
• * These prediction tools show potential for characterizing orphan proteins, especially those with known homologs or similar folds.
• * High-quality predictions for certain orphan proteins suggest the algorithms' utility in structural biology research.