Is it possible to also release the prompt used for each protein in protein_catalogue that used to generate the `generation`?

[qwu01]

Hey,
Congrats to the cool BioReason-Pro!
I was trying to obtain the SFT and RL generated texts for a subset of proteins (yeast to be specific) by running the predict.py, but the texts (all three sections: think, summary, and terms (interpro, GOBP/CC/MF terms) ) are different from what's released on hf-datasets protein_catalogue (terms are mostly the same, summary remains sematically similar, but is different).

Because the predicted go terms (go_pred) are released in huggingface training and testing dataset, I just used the released go_pred and bypassed the GO-GPT step. I only did GO-GPT for proteins that do not have go_pred data. This might be where the discrepancy from?
So I think it'd be best for reproducibility reasons that the prompt used for each protein could be released alongside the generation field of protein_catalogue.

Qi

[Adibvafa]

Thank you! This is valid feedback, I'll open-source the prompts too. That being said, i wouldn't expect a discrepency with the same protein. Could you let me know what is your data? I'll test to see why we get such discrepency. In particular im concerned with InterPro being different

[qwu01]

Hi,
Thanks, releasing the prompts would be very helpful.
To have more contexts, I changed the pyproject.toml line 18 from "torch>=2.6.0,<2.8.0", to "torch>=2.6.0,<2.10.0", in order to run, so my torch=2.9.1 being another difference. here's my venv:

virtualenv ../.bioreason_env_2
source ../.bioreason_env_2/bin/activate
cd BioReason-Pro
pip install --no-index --upgrade pip setuptools wheel packaging
pip install -e .
pip install esm --no-deps
pip install flash-attn --no-build-isolation --no-deps
python -c "import torch, flash_attn; print('torch:',torch.__version__); print('flash attention ok')"
pip install unsloth unsloth_zoo --no-deps
pip install zstd cloudpathlib tenacity biotite brotli msgpack msgpack-numpy biopython --no-deps

I focus on yeast (Saccharomyces cerevisiae (strain ATCC 204508 / S288c) (Baker's yeast)) proteins, for the interpro, yes I also used the hf-dataset column interpro_formatted.

for example, these two proteins: Q9ZZX9 and P41832, I used these prompt:

Q9ZZX9

    "prompt": "<|im_start|>user\nProtein: <|protein_pad|>\n\nGO graph: <|go_graph_pad|>\n\nYou are a scientific assistant specialized in protein function prediction. Given a protein sequence, organism information, and additional context (InterPro domain annotations and/or initial GO term speculations), step-by-step reason about the InterPro terms, Gene Ontology (GO) terms regarding molecular function, biological process, and cellular component, protein-protein interactions (PPI), and overall function. Use the provided information as a starting point and improve upon it with deeper analysis. Provide a summary of your findings in your final answer.\n\nGiven the protein above from organism Saccharomyces cerevisiae (strain ATCC 204508 / S288c) (Baker's yeast) with the following InterPro annotations:\nNone\n\nAnd the following initial GO term speculations:\nMolecular Function (MF): GO:0003674 (molecular function), GO:0003824 (catalytic activity), GO:0016740 (transferase activity), GO:0140096 (catalytic activity, acting on a protein), GO:0016757 (glycosyltransferase activity), GO:0016758 (hexosyltransferase activity), GO:0008194 (UDP-glycosyltransferase activity), GO:0008375 (acetylglucosaminyltransferase activity)\nBiological Process (BP): GO:0008150 (biological process), GO:0008152 (metabolic process), GO:0071704 (organic substance metabolic process), GO:0043170 (macromolecule metabolic process), GO:0010467 (gene expression)\nCellular Component (CC): GO:0005575 (cellular component), GO:0110165 (cellular anatomical entity), GO:0005622 (intracellular anatomical structure), GO:0043226 (organelle), GO:0005737 (cytoplasm), GO:0043229 (intracellular organelle), GO:0005739 (mitochondrion), GO:0043227 (membrane-bounded organelle), GO:0043231 (intracellular membrane-bounded organelle)\n\nReason about the function of the protein and focus more on its Molecular Function, Biological Process, Cellular Component. Summarize in UniProt format.<|im_end|>\n<|im_start|>assistant\n",

P41832

    "prompt": "<|im_start|>user\nProtein: <|protein_pad|>\n\nGO graph: <|go_graph_pad|>\n\nYou are a scientific assistant specialized in protein function prediction. Given a protein sequence, organism information, and additional context (InterPro domain annotations and/or initial GO term speculations), step-by-step reason about the InterPro terms, Gene Ontology (GO) terms regarding molecular function, biological process, and cellular component, protein-protein interactions (PPI), and overall function. Use the provided information as a starting point and improve upon it with deeper analysis. Provide a summary of your findings in your final answer.\n\nGiven the protein above from organism Saccharomyces cerevisiae (strain ATCC 204508 / S288c) (Baker's yeast) with the following InterPro annotations:\n- IPR016024: Armadillo-type fold (homologous_superfamily) [182-692]\n- IPR014768: Rho GTPase-binding/formin homology 3 (GBD/FH3) domain (domain) [174-696]\n- IPR015425: Formin, FH2 domain (domain) [1348-1824]\n- IPR011989: Armadillo-like helical (homologous_superfamily) [334-614]\n- IPR014767: Diaphanous autoregulatory domain (domain) [1792-1826]\n- IPR042201: Formin, FH2 domain superfamily (homologous_superfamily) [1412-1780]\n- IPR010472: Formin, FH3 domain (domain) [504-709]\n- IPR051661: Actin filament dynamics regulator (family) [46-1944]\n- IPR010473: Formin, GTPase-binding domain (domain) [173-459]\n\nAnd the following initial GO term speculations:\nMolecular Function (MF): GO:0003674 (molecular function), GO:0005488 (binding), GO:0005515 (protein binding), GO:0042802 (identical protein binding), GO:0005522 (profilin binding)\nBiological Process (BP): GO:0008150 (biological process), GO:0051179 (localization), GO:0048518 (positive regulation of biological process), GO:0050789 (regulation of biological process), GO:0022414 (reproductive process), GO:0040007 (growth), GO:0000003 (reproduction), GO:0065007 (biological regulation), GO:0009987 (cellular process), GO:0048519 (negative regulation of biological process), GO:0051641 (cellular localization), GO:0032505 (reproduction of a single-celled organism), GO:0007163 (establishment or maintenance of cell polarity), GO:0022402 (cell cycle process), GO:0032879 (regulation of localization), GO:0007049 (cell cycle), GO:0051234 (establishment of localization), GO:0051301 (cell division), GO:0071840 (cellular component organization or biogenesis), GO:0065008 (regulation of biological quality), GO:0051640 (organelle localization), GO:0019954 (asexual reproduction), GO:0050794 (regulation of cellular process), GO:0030029 (actin filament-based process), GO:0007117 (budding cell bud growth), GO:0007017 (microtubule-based process), GO:0048523 (negative regulation of cellular process), GO:0048522 (positive regulation of cellular process), GO:0032506 (cytokinetic process), GO:0007114 (cell budding), GO:0000910 (cytokinesis), GO:0044085 (cellular component biogenesis), GO:1903047 (mitotic cell cycle process), GO:0051129 (negative regulation of cellular component organization), GO:0030010 (establishment of cell polarity), GO:0060341 (regulation of cellular localization), GO:0051649 (establishment of localization in cell), GO:0051130 (positive regulation of cellular component organization), GO:0000226 (microtubule cytoskeleton organization), GO:0030036 (actin cytoskeleton organization), GO:0032970 (regulation of actin filament-based process), GO:0044087 (regulation of cellular component biogenesis), GO:0044837 (actomyosin contractile ring organization), GO:0051653 (spindle localization), GO:0051656 (establishment of organelle localization), GO:0016043 (cellular component organization), GO:0051128 (regulation of cellular component organization), GO:0000278 (mitotic cell cycle), GO:0090066 (regulation of anatomical structure size), GO:1902850 (microtubule cytoskeleton organization involved in mitosis), GO:0032535 (regulation of cellular component size), GO:0043254 (regulation of protein-containing complex assembly), GO:0043244 (regulation of protein-containing complex disassembly), GO:0000915 (actomyosin contractile ring assembly), GO:0032956 (regulation of actin cytoskeleton organization), GO:0032880 (regulation of protein localization), GO:0051294 (establishment of spindle orientation), GO:1902904 (negative regulation of supramolecular fiber organization), GO:0022607 (cellular component assembly), GO:0031333 (negative regulation of protein-containing complex assembly), GO:1902903 (regulation of supramolecular fiber organization), GO:0010639 (negative regulation of organelle organization), GO:0043242 (negative regulation of protein-containing complex disassembly), GO:0007015 (actin filament organization), GO:0061640 (cytoskeleton-dependent cytokinesis), GO:0030866 (cortical actin cytoskeleton organization), GO:1902410 (mitotic cytokinetic process), GO:0051293 (establishment of spindle localization), GO:0000912 (assembly of actomyosin apparatus involved in cytokinesis), GO:0033043 (regulation of organelle organization), GO:0006996 (organelle organization), GO:0031032 (actomyosin structure organization), GO:0010638 (positive regulation of organelle organization), GO:0097435 (supramolecular fiber organization), GO:0000281 (mitotic cytokinesis), GO:2000251 (positive regulation of actin cytoskeleton reorganization), GO:0032272 (negative regulation of protein polymerization), GO:0030835 (negative regulation of actin filament depolymerization), GO:0051493 (regulation of cytoskeleton organization), GO:0040001 (establishment of mitotic spindle localization), GO:0051494 (negative regulation of cytoskeleton organization), GO:2000249 (regulation of actin cytoskeleton reorganization), GO:1901879 (regulation of protein depolymerization), GO:0032271 (regulation of protein polymerization), GO:0030837 (negative regulation of actin filament polymerization), GO:0030832 (regulation of actin filament length), GO:1901880 (negative regulation of protein depolymerization), GO:1902407 (assembly of actomyosin apparatus involved in mitotic cytokinesis), GO:0045010 (actin nucleation), GO:1903475 (mitotic actomyosin contractile ring assembly), GO:0007010 (cytoskeleton organization), GO:0051017 (actin filament bundle assembly), GO:0061572 (actin filament bundle organization), GO:0110053 (regulation of actin filament organization), GO:0000132 (establishment of mitotic spindle orientation), GO:0051495 (positive regulation of cytoskeleton organization), GO:0008064 (regulation of actin polymerization or depolymerization), GO:0051693 (actin filament capping), GO:0030834 (regulation of actin filament depolymerization), GO:0030833 (regulation of actin filament polymerization), GO:0030865 (cortical cytoskeleton organization), GO:0030046 (parallel actin filament bundle assembly), GO:0051016 (barbed-end actin filament capping)\nCellular Component (CC): GO:0005575 (cellular component), GO:0032991 (protein-containing complex), GO:0110165 (cellular anatomical entity), GO:0005622 (intracellular anatomical structure), GO:0043226 (organelle), GO:0042995 (cell projection), GO:0016020 (membrane), GO:0030427 (site of polarized growth), GO:0005933 (cellular bud), GO:0071944 (cell periphery), GO:0060187 (cell pole), GO:0140535 (intracellular protein-containing complex), GO:0099080 (supramolecular complex), GO:0032153 (cell division site), GO:0005737 (cytoplasm), GO:0042763 (intracellular immature spore), GO:0000131 (incipient cellular bud site), GO:0051286 (cell tip), GO:0120025 (plasma membrane bounded cell projection), GO:0042764 (ascospore-type prospore), GO:0043332 (mating projection tip), GO:0043229 (intracellular organelle), GO:0005938 (cell cortex), GO:0005628 (prospore membrane), GO:0043228 (non-membrane-bounded organelle), GO:0005934 (cellular bud tip), GO:0005935 (cellular bud neck), GO:0005886 (plasma membrane), GO:0099081 (supramolecular polymer), GO:0043232 (intracellular non-membrane-bounded organelle), GO:0005937 (mating projection), GO:0099512 (supramolecular fiber), GO:0099513 (polymeric cytoskeletal fiber), GO:0005856 (cytoskeleton), GO:0015629 (actin cytoskeleton), GO:0005884 (actin filament)\n\nReason about the function of the protein and focus more on its Molecular Function, Biological Process, Cellular Component. Summarize in UniProt format.<|im_end|>\n<|im_start|>assistant\n",

Here are two protein examples (Q9ZZX9 and P41832) comparisons:
Q9ZZX9

"BioReason-Pro-SFT (Released)": "<think>
I begin with the InterPro architecture. The sequence carries a single catalytic module annotated as GO:0004252 serine-type endopeptidase activity. A solitary, well-defined serine protease domain implies a trimeric catalytic apparatus (His-Asp-Ser) embedded in one compact fold, without appended transmembrane helices or secretory targeting signals. This architecture causes endoproteolytic cleavage of polypeptide substrates rather than broad peptidoglycan or lipid hydrolysis, and it typically operates as a soluble or peripherally associated enzyme whose specificity is set by the local sequence around the catalytic serine and by transient binding surfaces in the domain.

From this catalytic capability, the most direct molecular function is serine-type endopeptidase activity (GO:0004252). A protease with this fold can cleave peptide bonds in nascent polypeptides. The stated role in promoting the formation of the first two N-linked glycans in the ER is mechanistically compatible with two non-exclusive scenarios that follow from the same catalytic logic: either the enzyme trims propeptides on secretory glycoproteins to expose or stabilize Asn-X-Ser/Thr sites for transfer, or it removes inhibitory segments on the oligosaccharyltransferase (OST) complex to maintain an active conformation. In both cases, endoproteolysis is the enabling chemistry.

These proteolytic events feed directly into protein N-linked glycosylation (GO:0006487). By generating the first glycans on nascent chains co-translationally, the enzyme ensures that the ER glycosylation machinery can act efficiently. If the protease acts on OST subunits, it would maintain the integrity and activity of the transferase complex; if it acts on substrates, it would increase the fraction of acceptor sequons rendered competent for glycan transfer. Either route causally elevates the efficiency of N-glycan installation.

Cellular localization follows from the process and the domain layout. The absence of transmembrane features and the involvement in co-translational N-glycosylation point to the endoplasmic reticulum environment, consistent with an ER-resident protease that engages secretory pathway clients. The ER is continuous with the nuclear envelope, which rationalizes detection at the nuclear envelope (GO:0005635) as part of the same membrane continuum. The presence of a DNA polymerase complex (GO:0042575) annotation can be reconciled by transient proximity at the nuclear periphery, where ER- and nucleus-adjacent protein assemblies can co-localize; a protease operating at the ER could encounter or regulate factors near the nuclear envelope without being a stable component of polymerase machinery. The mitochondrial matrix (GO:0005759) association is also plausible as a minor pool at ER–mitochondria contact sites, where quality control proteases sometimes sample or triage mistargeted polypeptides; the lack of transmembrane segments suggests any mitochondrial association would be peripheral or transient rather than lumenal.

The interaction set composed of multiple mitochondrial, putative uncharacterized proteins (Q0032, Q0017, Q0143, Q0297, Q0182, Q0092, Q0142) suggests a broader proteostasis network. A reasonable hypothesis is that the protease forms transient complexes with ER-associated factors that also touch mitochondrial surfaces, enabling surveillance of dual-targeted or mistargeted clients. Within the glycosylation axis, the protease likely associates with OST subunits and the Sec61 translocon to act co-translationally, while at the nuclear envelope it may encounter polymerase-associated scaffolds without direct catalytic action. Mechanistically, the enzyme’s serine protease domain would recognize exposed, unstructured segments of nascent chains or specific OST subunits, cleave a short inhibitory or propeptide segment, and thereby stabilize the active conformation of the transferase complex, increasing the probability and rate of the first N-glycan transfer in the ER.
</think>

- Functional Summary: A serine endopeptidase in baker’s yeast that operates in the endoplasmic reticulum to enhance the first co-translational N-glycan additions to nascent polypeptides. By proteolytically trimming inhibitory segments—either on substrates that become glycoproteins or on oligosaccharyltransferase subunits—it stabilizes the transferase machinery and increases glycan transfer efficiency. Its placement at the ER and contiguous nuclear envelope positions it near secretory entry points and nuclear-proximal assemblies, and a minor pool may act at ER–mitochondria contact sites. The enzyme’s core mechanism relies on a single catalytic serine protease domain that recognizes exposed peptide segments and executes precise endoproteolytic cleavages to promote glycosylation fidelity.
- UniProt Summary: Serine protease required for the formation of first two N-linked glycans in the endoplasmic reticulum.
- InterPro:
    - GO:0004252 serine-type endopeptidase activity
- Molecular Function:
    - GO:0004252 serine-type endopeptidase activity
- Biological Process:
    - GO:0006487 protein N-linked glycosylation
- Cellular Component:
    - GO:0005783 endoplasmic reticulum
    - GO:0005635 nuclear envelope
    - GO:0042575 DNA polymerase complex
    - GO:0005759 mitochondrial matrix",

    "BioReason-Pro-SFT (my run)": "<think>
I begin with the InterPro architecture. The sequence contains InterPro:0000030 (Carbohydrate-binding module 6), spanning residues 25–39. This is a compact lectin-like module that binds β-1,4-linked glucan chains. Its placement near the N-terminus suggests a luminal or extracellular-facing carbohydrate-recognition element that can dock the enzyme onto a growing polysaccharide chain. The protein is also described as a single-pass type II membrane protein, which implies an N-terminal cytosolic tail, a single transmembrane helix, and a luminal/extracellular C-terminal catalytic region. In this topology, the CBM6 module would lie in the same luminal/extracellular compartment as the catalytic center, positioning it to engage the nonreducing end of a β-1,4-glucan chain as it emerges during synthesis. This architecture causally supports a glycosyltransferase that extends a β-glucan polymer by transferring glucose from a UDP-glucose donor to the acceptor chain.

From this domain layout, the molecular function resolves to glucan 1,4-β-glucosyltransferase activity (GO:0003843). The type II membrane topology places the catalytic site in the secretory pathway lumen, where UDP-glucose is available via nucleotide-sugar transporters. The CBM6 module would increase local substrate affinity and processivity by tethering the enzyme to the nascent β-1,4-glucan, thereby promoting elongation of the chain. The same catalytic chemistry can also act on chitin to a lesser extent, consistent with a secondary chitin β-1,4-glucosyltransferase activity (GO:0004174), although the primary role is β-glucan elongation.

The biological process follows from the chemistry and location. By extending the outer wall β-1,4-glucan, the enzyme drives fungal-type cell wall biogenesis and remodeling, which underpins cell morphogenesis and growth. Strengthening and properly patterning the glucan matrix is essential for maintaining cell shape and for the integrity of the outer wall. Because cell wall assembly must be coordinated with transcriptional programs and mRNA metabolism that supply enzymes and regulators, the activity contributes to the broader concept of gene expression (GO:0010467) in the sense that the enzyme’s function is embedded in the gene-expression-driven production and deployment of wall-building machinery. Thus, the catalytic activity mechanistically links to transcriptionally regulated morphogenetic programs.

The cellular component is dictated by the topology. A single-pass type II membrane protein with a luminal CBM and catalytic domain is an integral component of membrane (GO:0016021), most plausibly the secretory pathway membranes (ER and Golgi) where cell wall polysaccharides are assembled and trafficked. The presence of a mitochondrion label (GO:0005739) suggests either a minor pool at ER–mitochondria contact sites or indirect association with mitochondrial processes that coordinate energy supply and lipid environment for secretory trafficking. While the catalytic role occurs on secretory pathway membranes, transient proximity to mitochondria or co-fractionation during cell fractionation could explain mitochondrial association signals.

Mechanistically, the enzyme likely operates as follows: the cytosolic N-terminus anchors the protein in the membrane; the luminal C-terminus contains the catalytic site that uses UDP-glucose as donor; the CBM6 (residues 25–39) binds the emerging β-1,4-glucan, aligning the nonreducing end for catalysis. Processivity and specificity are enforced by the CBM and local membrane microenvironment, ensuring efficient elongation of the outer wall glucan. Interaction partners are expected to include other cell wall synthases and regulators: a β-1,3-glucan synthase that provides the acceptor backbone; chitin synthase complexes that coordinate with glucan deposition; ER/Golgi nucleotide-sugar transporters that supply UDP-glucose; and scaffold proteins that organize synthase complexes at growth sites. These assemblies couple enzymatic elongation to morphogenetic cues, ensuring robust outer wall construction and maintenance.
</think>

- Functional Summary: A membrane-anchored glycosyltransferase in baker’s yeast that extends the outer wall β-1,4-glucan using UDP-glucose, with a luminal carbohydrate-binding module positioning the growing polymer for efficient catalysis. Its activity supports outer wall assembly and remodeling that underlie cell morphogenesis and growth, and it may also act on chitin to a lesser extent. The protein is a single-pass type II membrane enzyme functioning on secretory pathway membranes, with potential transient association near mitochondria–ER interfaces that coordinate energy and lipid environments for wall biogenesis.
- UniProt Summary: Beta-1,4-glucan synthase which catalyzes the elongation of the outer wall. May also have a chitin synthase activity. Required for outer wall assembly and remodeling, which are essential for cell morphogenesis and growth.
- InterPro:
    - GO:0000030 Carbohydrate-binding module 6
    - GO:0003843 glucan 1,4-beta-glucosyltransferase activity
    - GO:0004174 chitin beta-1,4-glucosyltransferase activity
- Molecular Function:
    - GO:0003843 glucan 1,4-beta-glucosyltransferase activity
    - GO:0004174 chitin beta-1,4-glucosyltransferase activity
- Biological Process:
    - GO:0010467 gene expression
- Cellular Component:
    - GO:0016021 integral component of membrane
    - GO:0005739 mitochondrion",

P41832

    "BioReason-Pro-SFT (Released)": "<think>
I begin with the InterPro architecture. A broad family signature, IPR051661 (Actin filament dynamics regulator, residues 46–1944), spans nearly the entire polypeptide, indicating a large, multidomain actin-assembly factor. The N-terminal half is dominated by a Rho-GTPase–responsive regulatory module: IPR010473 (Formin, GTPase-binding domain, residues 173–459) and IPR014768 (Rho GTPase-binding/formin homology 3 [GBD/FH3] domain, residues 174–696) overlap and extend into IPR010472 (Formin, FH3 domain, residues 504–709) with a second FH3 call at residues 506–702. These segments are built from helical repeats, as captured by IPR016024 (Armadillo-type fold, residues 182–692) and IPR011989 (Armadillo-like helical, residues 334–614). This helical scaffold forms the GBD/FH3 platform that binds Rho-family GTPases and mediates autoinhibition. A second GBD call (IPR010473, residues 298–457) sits within this region, consistent with a bipartite GTPase-binding surface. The C-terminal half contains the actin-assembly engine: IPR015425 (Formin, FH2 domain, residues 1348–1766) with overlapping calls IPR015425 (1348–1824) and IPR015425 (1348–1741), and a superfamily envelope IPR042201 (Formin, FH2 domain superfamily, residues 1412–1780). Immediately downstream lies IPR014767 (Diaphanous autoregulatory domain, residues 1792–1826), the DAD motif that docks onto the N-terminal GBD to enforce autoinhibition. The ordered layout—GBD/FH3 (armadillo-like helical scaffold) → (likely proline-rich FH1, though not explicitly annotated) → FH2 → DAD—creates a classic formin switch: the DAD binds the GBD to lock the FH2 in an inactive state; GTPase engagement at the GBD displaces DAD, freeing FH2 to nucleate and elongate actin filaments.

This architecture causes specific molecular activities. The FH2 dimer binds actin barbed ends and processively tracks them, which mechanistically explains barbed-end actin filament capping (GO:0051016) in the formin-specific sense—FH2 maintains a persistent cap that protects and directs elongation rather than allowing indiscriminate growth. The N-terminal proline-rich FH1 segment (implied by the formin family architecture between FH3 and FH2) recruits profilin–actin complexes, matching GO:0005522 profilin binding and enabling rapid monomer delivery to the FH2-bound barbed end. The large helical scaffold and the dimeric nature of FH2 promote self-association, supporting GO:0042802 identical protein binding. The GBD/FH3’s Rho-GTPase interface (and the FH3 targeting domain) couples the actin assembly engine to polarity cues, ensuring that actin assembly is spatially restricted.

From these molecular functions, the biological processes follow. FH2-driven nucleation (GO:0045010 actin nucleation) and processive elongation (GO:0030837 negative regulation of actin filament depolymerization) generate and maintain polarized actin cables, which in turn drive vectorial vesicle transport. This cable-based transport underlies budding cell apical bud growth (GO:0007118) and the establishment of mitotic spindle orientation (GO:0000132), because actin cables provide landmarks and forces that align the spindle. The same polarized actin assembly supports assembly of the mitotic actomyosin contractile ring (GO:1903475) by delivering actin and myosin machinery to the division site. The GTPase-regulated switch and cortical targeting also explain regulation of protein localization (GO:0032880), as the formin positions itself and its clients at the cortex. The ability of FH2 to remain as a persistent cap while elongating provides a mechanism for barbed-end actin filament capping (GO:0051016) that channels filament growth into bundles, enabling formin-nucleated actin cable assembly (GO:0070649).

The cellular context is dictated by targeting domains and the need to interface with membranes and actin. The FH3 domain targets the protein to polarity landmarks, consistent with localization at the incipient cellular bud site (GO:0000131), the cellular bud tip (GO:0005934), and later the cellular bud neck (GO:0005935). The cortical association and role in polarity place it at the cell cortex (GO:0005938) and plasma membrane (GO:0005886), where it organizes actin cables that emanate from the cortex. Its direct engagement with actin explains presence on the actin filament (GO:0005884). During mating, the same polarity machinery focuses at the mating projection tip (GO:0043332). During sporulation, cortical actin remodeling near the prospore membrane (GO:0005628) is supported by formin-mediated cable formation. The protein also forms a defined assembly with its partner scaffold, consistent with residence in the BUD6/BUD8 complex (GO:0031521).

These features suggest a mechanism: the GBD/FH3 (armadillo-like) region binds active Rho-family GTPases and adaptors at cortical polarity sites, displacing the DAD from the FH2 dimer. Freed FH2 nucleates a new actin filament, recruits profilin–actin via the intervening proline-rich segment, and remains as a processive cap to elongate the barbed end while maintaining filament alignment. This produces long, polarized actin cables that guide secretory vesicles to the bud tip, coordinate spindle orientation, and scaffold contractile ring assembly. The protein likely dimerizes via FH2 (identical protein binding), and its cortical targeting is stabilized by the BUD6/BUD8 complex.

From this mechanism, interaction partners can be hypothesized. The GBD/FH3 binds the small GTPase Cdc42 (Cell division control protein 42) and other Rho-family GTPases (GTP-binding protein of the rho subfamily of Ras-like proteins; GTPase-activating protein of the Ras superfamily; Rab GTPase-activating protein as part of polarity-linked trafficking). The FH1 segment recruits profilin–actin complexes (Profilin and Actin). The FH3 and BUD8 scaffold engage Bud site selection protein 6 (Bud6) and SPA2 and PEA2 to form the BUD6/BUD8 complex that targets the formin to polarity sites. The FH2 dimerization explains identical protein binding and association with the paralog BNI1-related protein 1, enabling functional redundancy and potential hetero-oligomers at distinct cortical zones.
</think>

- Functional Summary: A cortical, polarity-targeted actin assembly factor in baker’s yeast that uses an N-terminal GTPase-responsive regulatory module to relieve autoinhibition and activate a C-terminal actin-assembly engine. Once engaged at cortical polarity landmarks, it nucleates and processively elongates barbed ends while maintaining a persistent cap, thereby building long, aligned actin cables that direct vesicle traffic to the bud tip, coordinate spindle orientation, and scaffold the contractile ring. Recruitment by a polarity scaffold and activation by Rho-family GTPases ensure spatially restricted actin dynamics, and profilin-bound monomers accelerate filament growth at the barbed end.
- UniProt Summary: Required for the assembly of interphase actin cables. Acts redundantly with BNI1 to regulate the organization of the actin cytoskeleton and cell polarity. The two formins appear to have distinct roles in establishing and maintaining cell polarity. BUD7 is required for the formation of long actin cables that extend from the bud site to the bud tip and function in the transport of secretory vesicles to the site of bud growth. BUD7 is also involved in the correct positioning of the mitotic spindle and the assembly of the contractile ring.
- InterPro:
    - IPR051661: Actin filament dynamics regulator (family) [46-1944]
    - IPR010473: Formin, GTPase-binding domain (domain) [173-459]
    - IPR014768: Rho GTPase-binding/formin homology 3 (GBD/FH3) domain (domain) [174-696]
    - IPR016024: Armadillo-type fold (homologous_superfamily) [182-692]
    - IPR010473: Formin, GTPase-binding domain (domain) [298-457]
    - IPR011989: Armadillo-like helical (homologous_superfamily) [334-614]
    - IPR010472: Formin, FH3 domain (domain) [504-709]
    - IPR010472: Formin, FH3 domain (domain) [506-702]
    - IPR015425: Formin, FH2 domain (domain) [1348-1766]
    - IPR015425: Formin, FH2 domain (domain) [1348-1824]
    - IPR015425: Formin, FH2 domain (domain) [1349-1741]
    - IPR042201: Formin, FH2 domain superfamily (homologous_superfamily) [1412-1780]
    - IPR014767: Diaphanous autoregulatory domain (domain) [1792-1826]
- Molecular Function:
    - GO:0042802 identical protein binding
    - GO:0005522 profilin binding
- Biological Process:
    - GO:0007118 budding cell apical bud growth
    - GO:0032880 regulation of protein localization
    - GO:0030837 negative regulation of actin filament depolymerization
    - GO:0045010 actin nucleation
    - GO:1903475 mitotic actomyosin contractile ring assembly
    - GO:0000132 establishment of mitotic spindle orientation
    - GO:0051016 barbed-end actin filament capping
    - GO:0070649 formin-nucleated actin cable assembly
- Cellular Component:
    - GO:0031521 BUD6/BUD8 complex",

    "BioReason-Pro-SFT (my run)": "<think>
I begin with the InterPro architecture. The N-terminal half is dominated by a Rho GTPase-binding/formin homology 3 module: IPR014768 (GBD/FH3 domain, residues 174–696) contains IPR010473 (Formin, GTPase-binding domain, residues 173–459) and IPR010472 (Formin, FH3 domain, residues 504–709). This region is built from armadillo-like helical repeats, as indicated by IPR016024 (Armadillo-type fold, residues 182–692) and IPR011989 (Armadillo-like helical, residues 334–614). Such tandem armadillo repeats create a curved solenoid that binds Rho-family GTPases and mediates autoinhibitory contacts. The C-terminal half contains the actin assembly engine: IPR015425 (Formin FH2 domain, residues 1348–1824) within IPR042201 (Formin FH2 domain superfamily, residues 1412–1780). Immediately adjacent to the FH2 tip lies IPR014767 (Diaphanous autoregulatory domain, residues 1792–1826), the DAD motif that docks onto the N-terminal GBD/FH3 to impose autoinhibition. The entire polypeptide is classified by IPR051661 (Actin filament dynamics regulator family, residues 46–1944), consistent with a dedicated actin assembly factor. The linear order—N-terminal GBD/FH3 armadillo solenoid followed by a long proline-rich linker (implied between FH3 and FH2) and a C-terminal FH2 plus DAD—creates a classic formin switch: the DAD binds the N-terminus to lock the FH2 in an inactive state; binding of an active Rho-family GTPase to the GBD/FH3 relieves this clamp, exposing the FH2 to nucleate and elongate actin filaments.

This architecture causes two specific molecular activities. First, the FH2 dimer forms a processive barbed-end cap that both nucleates and tracks elongation, which mechanistically explains barbed-end actin filament capping (GO:0051016) and the ability to promote actin nucleation (GO:0045010) and positive regulation of actin cytoskeleton reorganization (GO:2000251). Second, the proline-rich segment between FH3 and FH2 recruits profilin–actin complexes, accounting for profilin binding (GO:0005522) and the observed stimulation of filament elongation. Formins also dimerize via FH2 interfaces, supporting identical protein binding (GO:0042802). Together, these activities drive actin filament bundle assembly (GO:0030046) and actin cross-link formation (GO:0051764) by producing long, unbranched filaments that can be cross-linked by cellular partners.

These molecular functions propagate to cellular processes. By generating long, polarized actin cables, the protein establishes directional transport tracks that underpin budding cell apical bud growth (GO:0007118). The same cables and the contractile ring’s actin structures enable mitotic actomyosin contractile ring assembly (GO:1903475) and its spatial control, regulation of protein localization to contractile ring actomyosin structure (GO:1904498). Actin cables also position microtubule-organizing centers, explaining establishment of mitotic spindle orientation (GO:0000132). The FH2-mediated barbed-end capping (GO:0051016) simultaneously limits branch nucleation, providing a mechanism for negative regulation of actin filament branching (GO:0051019).

The domain layout predicts a cortical, cortex-anchored localization. FH3/armadillo repeats often target polarity landmarks, and the FH2-driven actin assembly produces cortical cables and the contractile ring. This places the protein at the cell cortex (GO:0005938) and plasma membrane (GO:0005886) interface, and within the actin filament network (GO:0005884). Polarity targeting and function at growth sites explain presence at the incipient cellular bud site (GO:0000131), the cellular bud tip (GO:0005934), and the mating projection tip (GO:0043332). Its role in cytokinesis and septation accounts for localization at the cellular bud neck (GO:0005935). During sporulation, actin-driven membrane biogenesis and guidance rationalize presence at the prospore membrane (GO:0005628). The protein’s activity is also required for the formation of the actomyosin contractile ring, consistent with its FH2-dependent filament assembly and capping.

Mechanistically, I propose an autoinhibited dimer in which the DAD (residues 1792–1826) binds the N-terminal GBD/FH3 armadillo solenoid (residues 174–696). Activation occurs when a Rho-family GTPase engages the GBD, releasing DAD and enabling the FH2 dimer (residues 1348–1824) to nucleate filaments, cap barbed ends processively, and recruit profilin–actin via the intervening proline-rich region to accelerate elongation. The resulting cables guide myosin-driven transport to bud tips and mating projections, and scaffold the contractile ring at the bud neck. Likely interaction partners include the Rho GTPase Cdc42 bound to the GBD/FH3, profilin (e.g., Asn1) bound to the proline-rich region, actin (Act1) as substrate, formins such as Bni1 for cooperative cable assembly, and contractile ring components (Myo1, Rlc1, Cdc15, Mid1) that couple filament assembly to ring assembly and cytokinesis. These interactions collectively explain the observed roles in polarized growth, spindle orientation, and contractile ring formation.
</think>

- Functional Summary: A cortical actin assembly factor in baker’s yeast that uses an autoinhibited formin switch to nucleate and elongate actin filaments. Binding of an activating Rho-family GTPase to its N-terminal regulatory solenoid releases a C-terminal dimeric actin-assembly core that both nucleates and processively caps barbed ends while recruiting profilin–actin to accelerate elongation. By generating long, unbranched actin cables and contributing to the contractile ring, it directs polarized growth at the bud tip and mating projection, positions the mitotic spindle, and organizes protein localization at the actomyosin ring. It operates at the plasma membrane–proximal cortex, including the incipient bud site, bud tip, bud neck, and sporulation structures, to coordinate actin filament formation, bundling, and cross-linking with spatial precision.
- UniProt Summary: Required for the formation of the actomyosin contractile ring. Nucleates actin filaments and stimulates their elongation. Stimulates actin filament bundling and actin cross-link formation.
- InterPro:
    - IPR016024: Armadillo-type fold (homologous_superfamily) [182-692]
    - IPR014768: Rho GTPase-binding/formin homology 3 (GBD/FH3) domain (domain) [174-696]
    - IPR015425: Formin, FH2 domain (domain) [1348-1824]
    - IPR011989: Armadillo-like helical (homologous_superfamily) [334-614]
    - IPR014767: Diaphanous autoregulatory domain (domain) [1792-1826]
    - IPR042201: Formin, FH2 domain superfamily (homologous_superfamily) [1412-1780]
    - IPR010472: Formin, FH3 domain (domain) [504-709]
    - IPR051661: Actin filament dynamics regulator (family) [46-1944]
    - IPR010473: Formin, GTPase-binding domain (domain) [173-459]
- Molecular Function:
    - GO:0042802 identical protein binding
    - GO:0005522 profilin binding
- Biological Process:
    - GO:0007118 budding cell apical bud growth
    - GO:2000251 positive regulation of actin cytoskeleton reorganization
    - GO:1904498 regulation of protein localization to contractile ring actomyosin structure
    - GO:0045010 actin nucleation
    - GO:1903475 mitotic actomyosin contractile ring assembly
    - GO:0000132 establishment of mitotic spindle orientation
    - GO:0051019 negative regulation of actin filament branching
    - GO:0051764 actin cross-link formation
    - GO:0030046 parallel actin filament bundle assembly
    - GO:0051016 barbed-end actin filament capping
- Cellular Component:
    - GO:0000131 incipient cellular bud site
    - GO:0043332 mating projection tip
    - GO:0005938 cell cortex
    - GO:0005628 prospore membrane
    - GO:0005934 cellular bud tip
    - GO:0005935 cellular bud neck
    - GO:0005886 plasma membrane
    - GO:0005884 actin filament",