Fig 1.
Fractional coverage of proteomes by DPAM classification.
DPAM classifies regions of AlphaFold2 models into one of six categories: 1) Assigned domains (blue) are globular domains with well-defined homologous links to a reference domain 2) Unassigned domains (yellow) are globular with well-defined secondary structure but a confident automated link to a reference domain could not be determined. 3) Flexible linkers / disordered regions (gray) that are determined by PAE to be non-globular and in low-confidence interactions with other residues. 4) Partial domains (cyan) possess high-confidence links to reference domains where no high-coverage alignment can be found. 5) Simple domains (green) contain 2 or fewer secondary structure elements. 6) Low confidence regions (orange) are regions that cannot be assigned to domains and have low (<70 plDDT) confidence. Organisms are ordered by their NCBI taxonomy identifier from a tree built with the NCBI Taxonomy CommonTree tool such that organisms with the same domain, phylum, genus, etc. are grouped [34].
Fig 2.
Most populated ECOD groups of DPAM domains in AlphaFold DB whole proteomes.
A) Relative populations of DPAM domains assigned to ECOD by class (inner pie) and architecture (outer donut). B) Twenty most populated ECOD homologous groups from all AF2 whole proteomes, colored by class. C) An uncharacterized maize CTPR repeat (AFDB: A0A1X7YGI2). Internal repeat proteins (such as CTPR) can be subdivided into smaller repeated domains(d1:red, d2:blue, d3:green). D) Rice disease-related protein 1 (DRP1, AFDB: Q84QL4). The HTH domain of this protein (purple) is shown in the context of other detected domains. E) Fibronectin-binding protein B (FnbpB) of S. aureus (AFDB: Q2G1T5).
Fig 3.
Comparison of DPAM domain population in eukaryotes and bacteria from 48 AF2 DB proteomes.
A) Top 20 most populated homologous groups in domains classified from AFDB eukaryotes colored by structure class B) Top 20 most populated homologous groups in domains classified from AFDB bacteria colored by structure class C) Most populated homologous groups (>0.5% domain population) domain population stratified by species and normalized by total species and homologous group population(i.e. row and column).; D) B. Malayi C2H2-type domain-containing protein (AFDB: A8PDW8) contains three (d1:red, d2:blue, d3green) repeated well-defined BBA zinc finger domains, as well as four ZnF repeats that were filtered for lacking sufficient topology (magenta). E) C. elegans RING finger-containing protein (AFDB: Q93343) contains one well-assigned RING finger ZnF domain (d1,red), along with HTH (d2, blue), and BBA Znf (d3, green) domains, also show filtered simple topology domains (magenta) along with a low-pLDDT region (cyan) F) C. elegans (AFDB: O45201) Delta-like protein dsl-1 contains one well-assigned EGF domain (d1, red) along with one immunoglobulin-like domain (d2, blue) as well as a EGF-domain with poorly formed topology (magenta).
Fig 4.
DPAM domain classification compared to ECOD experimental reference.
A) Top 20 most populated homologous groups in ECOD v285 reference set, composed entirely of experimentally determined proteins and their domains B) Non-redundant (F70) domain population of ECOD v285 and v25 (i.e., 7 years earlier). C) Relative domain population of ECOD F70 eukaryotic representative domains compared to relative domain population of AF2 eukaryotic domains. D) Relative domain population of ECOD F70 bacterial representative domains compared to relative domain populations of AF2 bacterial domains.
Fig 5.
Major facilitator superfamily domains in experimental and predicted structures.
A) E. coli lactose permease (PDB: 1PV7) and its two defined ECOD domains: e1pv7A1 (red), e1pv7A (blue).2). B) PGAP2-interacting protein (UNP: Q9H720) with two previously unannotated MFS domains (red, blue) and a known DNase I-like (cyan) domain C) solute carrier family 45 member 3 (UNP: Q96JT2) contains an MFS domain pair (red, blue) with a large insertion in the C-terminal domain. The annotation of this C-terminal domain is truncated in InterPro annotations.
Fig 6.
Orphan domain Duf3676 represents a fast-evolving killer toxin-like fold.
Core fold common to Duf3676 (A-C) and yeast killer toxins (D-F) are depicted in rainbow cartoon from blue (N-terminus) to red (C-terminus). Cysteine side chains (gray stick) mark the disulfides, and termini are labeled. A) Duf3676 CSS2 AF model has an N-terminal extension (slate) to the common core fold that is decorated by an edge strand from the C-terminus (magenta). B) Duf3676 representative YER076C colored as in A, with the C-terminal strand in an opposite orientation. C) Duf3676 domain with an N-terminal strand (slate) replacing the variable edge strand. D) Top identified DPAM domain (ECOD: e6greB2) from Duf26 Gnk2 homolog has a similar N-terminal edge strand (slate) as Duf3676 domain in panel C, but alternate disulfide patterns (gray stick). E) Gnk2 binds carbohydrate (black stick) near the N-terminal strand. F) Yeast killer toxin bound to carbohydrate (black stick).