Generated by write_results.py from results.json; results.json is
produced by nucleotide_site.py. Do not hand-edit — a fresh run will
overwrite it.
uv run python nucleotide_site.py
uv run python write_results.py
UniProt places ACTR5 in "the actin family, ARP5 subfamily" and CDD
annotates it cd10211 ASKHA_NBD_Arp5, i.e. an actin/Hsp70/sugar-kinase
nucleotide-binding domain. A fold assignment is not an activity, so two
things are tested here, in both directions:
| PDB | resolution (A) | ARP5 chain | nucleotide in the ARP5 chain | contacts (<= 4.0 A) | ATP-mimic group in the ARP5 chain |
|---|---|---|---|---|---|
| 6HTS | 4.8 | H | none | - | - |
| 7ZI4 | 3.2 | H | ADP | 16 | none |
| 9GCG | 3.43 | J | ADP | 14 | none |
| 9GE5 | 3.35 | J | ADP | 12 | none |
| 9GEV | 3.47 | J | none | - | - |
| 9GFB | 3.55 | J | none | - | - |
Resolutions are the PDBe-reported values, fetched rather than transcribed.
ARP5 chain assignment comes from the PDBe SIFTS UniProt mapping, not from
chain letters: a missing Q9H9F9 mapping aborts the run rather than
reporting an empty result (a silent zero reads as a finding).
ADP is modelled in the ARP5 chain in 3 of 6 entries (7ZI4, 9GCG, 9GE5); ATP never is.
The three entries without it are 6HTS (whose ARP5 was an I-TASSER homology
model trimmed of all side chains) and two of the 2026 nucleosome states.
The bound species is plain ADP, not an ATP analogue. Every one of these
samples was soaked with ADP-BeF3, which would ordinarily leave the
identity of the ligand ambiguous. It does not here: in all three entries
the ARP5 chain contains no BeF3, AlF, VO4 or PO4 group alongside its ADP.
* 7ZI4: mimic groups in the ARP5 chain: none; elsewhere in the entry: BEF in chain G.
* 9GCG: mimic groups in the ARP5 chain: none; elsewhere in the entry: none in the entry.
* 9GE5: mimic groups in the ARP5 chain: none; elsewhere in the entry: none in the entry.
In 7ZI4 the only BeF3 in the whole entry sits in chain G, the Ino80 motor
domain, which is where the ATP mimic belongs. So the ADP-BeF3 soak
qualifies what the motor was trapped with, not what ARP5 was: ARP5's
cleft holds ADP. What the structures still cannot do is rank ADP against
ATP, because no ATP was offered to ARP5 in solution.
Contacts in the highest-resolution entry (7ZI4, 3.2 A):
G38 S39 F40 R43 D166 G189 Y190 G215 S216 E243 H247 G496 N497 M499 Y500 V530
Reference: 2BTF chain A (ACTB_BOVIN, P60712, 100.0% identical to human P60709), ATP-contacting residues:
G13 S14 G15 M16 K18 G156 D157 G158 V159 G182 K213 E214 G301 G302 T303 M305 Y306 K336
Finding a ligand is not enough — it has to be the same site. So the
test is reciprocal: align ACTR5 to beta-actin, then ask what fraction of
ARP5's observed contacts fall on ACTR5 positions that align to
beta-actin's own nucleotide contacts.
| PDB | observed ARP5 contacts | on the aligned actin pocket | fraction | off-pocket |
|---|---|---|---|---|
| 7ZI4 | 16 | 13 | 0.81 | 166, 216, 243 |
| 9GCG | 14 | 11 | 0.79 | 101, 216, 243 |
| 9GE5 | 12 | 8 | 0.67 | 37, 135, 166, 188 |
Union over the three entries: 20 ACTR5 positions contact the nucleotide, 13 of them on the aligned actin pocket. The retained contacts are the two actin phosphate-binding
loops (ACTB G13-S14-G15-M16-K18 -> ACTR5 G38-S39-F40-Q41-R43; ACTB G156
-> ACTR5 G189) and the adenosine shelf (ACTB G302/M305/Y306 -> ACTR5
G496/M499/Y500). This is the actin nucleotide cleft, not an adventitious
surface site.
Literature-defined actin catalytic positions (ACTB numbering), kept
deliberately identical to the ACTL7A review's audit so the two are
comparable: D11, Q137, D154, V159, H161.
The script asserts each position carries the expected residue in ACTB
before mapping anything.
ACTR5 = DDSCH, i.e. 2/5 conserved: the
Mg-coordinating Asp11 and His161 are kept, but Gln137 (which orients the
attacking water), Asp154 and Val159 are not. So the pocket binds a
nucleotide while the hydrolysis machinery is degenerate — the same shape
of result the ACTL7A review reported for its gene.
Cross-check: the independently committed human/ACTL7A/ACTL7A-bioinformatics/results.json reports DDSCH for ARP5_HUMAN; this run computes DDSCH (agree = True). A mismatch is a hard error, not a warning.
Reference: 8A2S (ACTS_RABIT, P68135, 93.6% identical to human beta-actin), chains A,B,C,D,E. Interface = residues within 4.5 A of a different protomer in >= 50% of protomers that have any inter-protomer contact: 72 positions in ACTB numbering.
| protein | group | catalytic | nucleotide pocket %id | protomer interface %id | overall %id vs ACTB |
|---|---|---|---|---|---|
| ACTB_HUMAN | conventional_actin | DQDVH (5/5) | 100.0 | 100.0 | 100.0 |
| ACTA1_HUMAN | conventional_actin | DQDVH (5/5) | 94.4 | 94.4 | 93.6 |
| ARP1_ACTR1A_HUMAN | filament_forming_arp | DQDVH (5/5) | 72.2 | 51.4 | 53.6 |
| ARP3_ACTR3_HUMAN | nucleotide_binding_arp | DQDVH (5/5) | 72.2 | 30.6 | 41.2 |
| ARP11_ACTR10_HUMAN | divergent_arp | DSDEL (2/5) | 50.0 | 22.2 | 30.1 |
| ARP4_ACTL6A_HUMAN | nuclear_arp | DTDHT (2/5) | 44.4 | 36.1 | 40.5 |
| ARP6_ACTR6_HUMAN | nuclear_arp | DADFH (3/5) | 44.4 | 18.1 | 29.3 |
| ARP8_ACTR8_HUMAN | nuclear_arp | HEDKS (1/5) | 55.6 | 22.2 | 24.5 |
| ARP5_HUMAN | query | DDSCH (2/5) | 44.4 | 20.8 | 32.5 |
| ARP5_DROME | arp5_ortholog_iba_donor | DDSTH (2/5) | 33.3 | 16.7 | 27.0 |
| ARP5_YEAST | arp5_ortholog_iba_donor | DDSDN (1/5) | 38.9 | 19.4 | 31.5 |
| ARP5_SCHPO | arp5_ortholog_iba_donor | DDNAH (2/5) | 33.3 | 22.2 | 30.6 |
| ARP5_ARATH | arp5_ortholog_iba_donor | DDCTH (2/5) | 33.3 | 22.2 | 32.6 |
ARP1/ACTR1A is the positive control: it is the one ARP that really does
form a filament, and it keeps 51.4% of the
interface, against 20.8% for ACTR5 — a
2.5-fold
gap, so the measurement discriminates. ACTR5 sits with the
non-polymerising nuclear ARPs.
Robustness: repeating the whole transfer under BLOSUM45 with a
(-14,-2) gap model gives ACTR5 pocket 44.4% / interface 18.1% and the same
catalytic string DDSCH, so the
conclusions do not depend on the alignment model.
The four WITH/FROM donor orthologs (fly, budding yeast, fission yeast,
Arabidopsis ARP5) all show a partly conserved nucleotide pocket
(DROME 33.3%, YEAST 38.9%, SCHPO 33.3%, ARATH 33.3%)
and a degenerate filament interface
(DROME 16.7%, YEAST 19.4%, SCHPO 22.2%, ARATH 22.2%).
So both results are family-wide properties of ARP5, not human
peculiarities: any nucleotide-binding gap in the GO record is a
family-level gap, and the absence of filament terms across the family is
correct rather than an oversight.
primaryAccession is the one requested).structures/ and are not