Kuiz berkaitan panduan · Keras Tahap
Kimia Generatif dan Kuiz Reka Bentuk Molekul De Novo
Fahami cara kimia generatif mencadangkan calon dan menilai kesahan kimia, objektif, kebaharuan, kebolehsintesisan dan susulan eksperimen.
soalan 1 daripada 8
What does de novo molecular design ask a model to do?
Pratonton setiap soalan8 soalan
Jawapan dan penjelasan muncul semasa anda mengambil kuiz.
- What does de novo molecular design ask a model to do?
- Only classify known compounds
- Measure a clinical endpoint directly
- Propose new molecular structures under stated objectives
- Replace chemical synthesis
- Mengapa model boleh memanfaatkan fungsi ganjaran molekul?
- It always learns the true biological mechanism
- It may find structures that score well because the predictor has blind spots
- Reward scores are direct experimental measurements
- Chemical constraints guarantee activity
- Apa yang ditetapkan oleh graf molekul yang sah?
- That the compound binds the target
- That it can be synthesized at scale
- That clinical efficacy is proven
- That the structure follows encoding rules, not that it is useful or safe
- Bagaimana kebaruan harus ditafsirkan dalam laporan kimia generatif?
- Relative to a specified and dated reference set after standardization
- As proof that a molecule has never existed anywhere
- Only by counting generated strings
- Without considering salts or stereochemistry
- Mengapa menyertakan ulasan kebolehlaksanaan sintesis?
- Models can rank commercially available reagents without route planning
- Synthesis feasibility proves the predicted target activity
- A predicted property score does not show that a route is practical
- A high reward guarantees the structure can be made at scale
- What can go wrong when optimizing only a predicted activity score?
- It guarantees molecular diversity
- The predictor may be wrong outside its training domain
- It prevents reward hacking
- It measures the real assay directly
- Which evaluation set design better tests chemical generalization?
- Random duplicate structures across every split
- The training set alone
- Only candidates with the highest reward
- A scaffold-separated held-out set
- What ultimately establishes whether a proposed candidate has the desired biological behavior?
- Experimental assays and expert interpretation
- A validity filter alone
- The model architecture name
- A high generative likelihood