一键导入
mutation-design-gfp
Propose high-fluorescence and high-diversity mutants of Green Fluorescent Protein (GFP) through multi-round iterative optimization.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
菜单
Propose high-fluorescence and high-diversity mutants of Green Fluorescent Protein (GFP) through multi-round iterative optimization.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Find the most suitable skill for a given biomedical task. Use this skill when: (1) You are unsure which skill to use for a specific biomedical task, (2) You want to discover available skills for a particular domain, (3) You need to compare multiple skills for a given use case.
Generate diverse druggable molecules for a given target or disease using OpenBioMed's AI-powered drug discovery tools. Use this skill when: (1) Generating drug candidates, molecules, or compounds for a target/disease, (2) Performing structure-based drug design or de novo drug design, (3) Finding or creating molecules that bind to a specific protein target, (4) Discovering potential drugs for a disease name, (5) Designing molecules with specific properties (LogP, QED, docking scores). The skill handles target identification, structure retrieval, molecule generation, and in silico evaluation.
Analyze potential drug-drug interactions (DDI) for up to 5 drugs using KEGG DDI database. Use this skill when: (1) Checking interactions between multiple medications, (2) Assessing DDI risk for drug combinations, (3) Understanding interaction mechanisms and severity, (4) Analyzing CYP enzyme involvement in DDIs.
Generate comprehensive drug development progress reports for disease therapeutic targets. Use when user asks about target drug pipeline, clinical trials, or research progress. Triggers on phrases like "target report", "drug development progress", "clinical trial summary", "靶点报告", "药物研发进展", "竞品分析", "专利分析".
Create new biomedical skills or improve existing ones for the OpenBioMed toolkit. Use this skill when: (1) Creating a new skill from scratch, (2) Capturing a workflow as a reusable skill, (3) Automating a biomedical task, (4) Improving an existing skill. This skill guides through an interactive process: define intent → design workflow → validate with real data → iterate → evaluate.
Query ChEMBL database for bioactivity data on drug-like compounds. Use this skill when: (1) Finding compounds active against a protein target (target-based search), (2) Getting bioactivity profile for a molecule (molecule-based search), (3) Finding drugs for a disease indication (indication-based search).
| name | mutation-design-gfp |
| description | Propose high-fluorescence and high-diversity mutants of Green Fluorescent Protein (GFP) through multi-round iterative optimization. |
A skill performs automated multi-round optimization of Green Fluorescent Protein (GFP) to discover mutants with higher fluorescence intensity and higher diversity.
Example prompts:
This skill can:
Download initial GFP sequences from https://cloud.tsinghua.edu.cn/f/5e673c1db710466b828f/?dl=1 and use them as the starting pool.
Download the oracle GFP prediction model from https://cloud.tsinghua.edu.cn/f/f655f79d7bb04a98a0bb/?dl=1, and the configuration file from https://cloud.tsinghua.edu.cn/f/8a894bb4b41f4074b9b0/?dl=1.
Execute code for oracle loading and scoring:
import torch
from omegaconf import OmegaConf
# ===== ORACLE MODEL LOADING =====
def load_oracle_model(ckpt_path, cfg_path):
with open(cfg_path, 'r') as fp:
cfg = OmegaConf.load(fp.name)
oracle = BaseCNN(**cfg.model.predictor)
state_dict = torch.load(ckpt_path)
oracle.load_state_dict(torch.load(ckpt_path))
oracle.eval()
# ===== ORACLE SCORING FUNCTION =====
def score_sequence(oracle, sequence: str) -> float:
results = oracle(sequence).detach()
return results.cpu().numpy()
Compute ESM2 embeddings for all sequences to represent sequence features.
Proposal: for each round, propose 96 × 4 candidate mutants from the current population using only point mutations with ≤4 mutations per sequence.
Evaluation: evaluate all candidate sequences using the oracle scoring function. Use oracle feedback from previous rounds to bias mutation proposals toward directions that increase predicted fluorescence (fitness gradient exploitation).
Selection: rank sequences by predicted fitness and select the top 96 mutants, while maintaining diversity measured by average pairwise Hamming distance.
Repeat proposal, evaluation, and selection until 10 rounds are completed, or best fitness does not improve for 3 consecutive rounds.
Collect the best 96 mutants discovered across all rounds and sort them by predicted fluorescence, and export the results as a CSV file following the specified output format.
The final result must be a CSV file with two columns:
| sequence | fitness |
|---|---|
| GFP_mutant_sequence | predicted_fluorescence |
Requirements:
Example:
sequence,fitness
SKGEELFTGVVPILVELDGDVNGHKFSVSGEGEGDATYGKLTLKFICTT...,0.93
SKGEELFTGVVPILVELDGDVNGHKFSVSGEGEGDATYGKLTLKFIATT...,0.91
SKGEELFTGVVPILVELDGDVNGHKFSVSGEGEGDATYGKLTIKFICTT...,0.89
...
This CSV represents the final optimized GFP mutant library predicted to exhibit higher fluorescence intensity.