Now 5224 visitors
Today:0 Yesterday:0
Total: 147749
212S 50P 4R
2026-10-20, Week 43 |
| Member Login |
| TACT Journal Homepage |
| Call for Paper |
| Committee |
| Find My Paper |
| Author Homepage |
| Statistics & History |
| Welcome Message |
| Paper Procedure |
| FAQ |
| Paper Archives |
| Outstanding Papers |
| Author Homepage |
| - Paper Procedure |
| - Journal Procedure |
| - Presentation Tips |
|
| Program & Proceedings |
| Program with Papers |
| Plenary Session |
| Tutorial Session |
|
| Presentation Platform |
| Hotel & Travel Info |
| About Korea |
| Accommodation |
| Transportation |
| VISA |
| Other Infomation |
|
| Photo Gallery |
| Scheduler Login |
| Seminar |
| Archives Login |
|
| Sponsors |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| Save Q&A |
 |
* Edit or answer any Q&A by selecting Q&A number Hyper Link below + Write button (Save)
|
ICACT20260263 Question.1
Questioner: mrajhans@paloaltonetworks.com 2026-02-11 오전 10:21:17 |
ICACT20260263 Answer.1
Answer by Auhor neyakhan57@gmail.com 2026-02-11 오전 10:21:17
Chrome  Click!! |
| How does the proposed structure-aware RAG system generalize to unseen CYP isoforms or proteins with sparse or lower-confidence structural data, given that evaluation is limited to three well-studied enzymes (CYP3A4, CYP2D6, CYP2C19)? |
Although the experimental evaluation was conducted on CYP3A4, CYP2D6, and CYP2C19, the proposed structure aware RAG framework is architecturally isoform agnostic and therefore generalizable to unseen CYP enzymes. The geometric embedding module encodes fundamental structural descriptors such as Cα spatial coordinates, inter residue distance distributions, secondary structure elements, and functional region types that are conserved across the CYP superfamily. Because these representations do not depend on isoform specific parameters, any CYP enzyme with available structural coordinates can be processed using the same embedding pipeline.
In addition, biology aware chunking is based on conserved functional domains including the heme binding region and substrate recognition sites, which enables alignment of structurally analogous regions across different isoforms. The cross enzyme attention mechanism further supports transferable knowledge representation by modeling shared structural patterns while preserving enzyme specific characteristics, as reflected by the cross enzyme retrieval performance.
For isoforms with sparse or lower confidence structural data, the hybrid retrieval formulation with alpha equal to 0.7 ensures robustness by balancing semantic and geometric similarity. When structural precision is limited, such as in predicted models, textual similarity maintains retrieval stability. Moreover, geometric features derived from distance distributions and cavity level properties are relatively tolerant to moderate coordinate uncertainty.
Although broader empirical validation across additional CYP isoforms is required, the architecture itself is scalable and does not require structural modification to incorporate new enzymes provided that structural coordinates are available. |
| |
