10. LLMs for Chemistry

Large Language Models, or LLMs, are advanced machine learning models capable of understanding and generating human-like text and answering QA questions based on extensive training data from diverse sources.

In chemistry, these models assist researchers and students by automating information retrieval, predicting chemical reactions, and supporting molecular design tasks. Benchmarks such as ChemLLMBench provide datasets to evaluate the accuracy and reliability of LLMs in chemical contexts.

So what exactly are LLMs and why are they so powerful? We will dive into this in this section.

10.1 Introduction of LLMs

What are LLMs?

• Large Language Models (LLMs) are smart computer systems like ChatGPT and Gemini. They learn from huge amounts of text data and can understand and generate human-like writing.

What can they do?

• These AI models can understand the meaning behind words, answer questions, and even come up with new ideas.
• In chemistry, LLMs are helpful because they can:
• Simplify complicated chemical information.
• Make difficult tasks easier to handle.
• Quickly find and organize important data.
• Help automate tasks that might otherwise take scientists a long time to complete.
• Summarize extensive research articles or reports succinctly.

In short, LLMs help make complex information clearer and easier to work with, even if you don’t have a chemistry background.

→ Such applications can substantially speed up research processes, improve accuracy, and open new avenues for innovative scientific discoveries.

Section 10.1 - Quiz Questions

1) Factual Questions

Question 1)

LLM is an acronym for:
A. Limited‑Length Memory
B. Large Language Model
C. Layered Lattice Matrix
D. Linear‑Loss Machine

Question 2)

Which capability is not typically associated with an LLM used in chemistry?
A. Summarising a 50‑page synthetic protocol
B. Translating SMILES to IUPAC
C. Running ab initio molecular‑orbital calculations
D. Answering literature questions in natural language

2) Comprehension / Application Question

Question 3)

You ask an LLM: “List three advantages of using ChatGPT when preparing a lab report.”
Which response below shows that the model has understood the chemistry context rather than giving generic writing advice?
A. Use active voice, avoid jargon, add headings.
B. Quickly draft the experimental Procedure section, check safety phrases (e.g., ‘corrosive,’ ‘flammable’), and suggest ACS‑style citations for each reagent.
C. Provide colourful images, add emojis, write shorter paragraphs.
D. Increase your font size to 14 pt.

10.2 Prompt Engineering

Definition:

Prompt engineering involves carefully writing instructions (prompts) to guide LLMs to give accurate and useful answers. In simpler terms, it’s how you ask questions or provide information to an AI system clearly and specifically.

Why is it important in chemistry?

Chemists often ask AI models to help with complex tasks, such as predicting chemical reactions or interpreting experimental data. Clear and detailed prompts help the AI understand exactly what information you’re looking for.

Format of a Good Prompt:

1. General Instruction: Briefly describes the task context or role the model is expected to perform.
   Example: “You are an expert chemist. Given the [Input Representation]: [Input Data], predict the [Target Output] using your domain knowledge. Provide [Input Explanation], [Output Explanation & Restrictions], and [Few-shot examples].”

2. Input Representation: Type of input (e.g., SMILES, molecular description).
   Example: “SMILES string”, “molecular structure”.

3. Input Data: The specific data instance the model will work with.
   Example: CC(=O)OC1=CC=CC=C1C(=O)O (Aspirin) or “benzene + nitric acid”.

4. Target Output: Expected prediction/result.
   Example: “product SMILES”.

5. Relevant Domain: The chemical expertise area to use.
   Example: “organic chemistry”.

6. Input Explanation: Clarifies provided data.
   Example: “Given the reactants aspirin and hydrochloric acid…”

7. Output Explanation & Restrictions: Desired response format and limits.
   Example: “Provide the reaction products in SMILES notation only.”

8. Few-shot Examples: Demonstrate the expected format. Example: Input: Reactants: aspirin + hydrochloric acid
   Output (SMILES):

Input: Reactants: benzene + nitric acid
Output (SMILES):

Example Task and Prompt:

Task: Analyze the interaction between hydrogen gas and water.
“You are an expert chemist. Explain what happens when hydrogen gas (H₂) is introduced to water (H₂O) under standard conditions. Include:

• Physical and chemical properties of hydrogen in water
• Possible reactions or interactions
• Conditions for reaction
• Energy changes or hazards
• Explanation of why the reaction proceeds or not under normal conditions

Provide relevant safety considerations.
Example: Analyze the interaction between chlorine gas and water.
Answer:
Chlorine gas is a yellow‑green diatomic molecule with moderate solubility in water (≈0.7 g/L at 20 °C) and a strong oxidizing character. When Cl₂ dissolves, it undergoes a hydrolysis equilibrium:Cl₂(g) + H₂O(l) ⇌ HCl(aq) + HOCl(aq). Under ambient conditions this reaction is mildly exothermic, releasing enough heat to warm the solution but not boil it. Hydrochloric acid and hypochlorous acid lower the pH, driving further Cl₂ uptake until equilibrium is reached. Hypochlorous acid is an effective disinfectant, owing to its oxidizing power. The reaction proceeds spontaneously because Cl₂ is a strong electrophile and water acts as a nucleophile, facilitating bond cleavage. Hazards include toxicity and corrosiveness of both gaseous Cl₂ and its acidic products; HOCl can bleach skin or materials. Always handle chlorine in a fume hood, wear acid‑resistant gloves, goggles, and respiratory protection, and neutralize any acidic effluent before disposal.”

Tips:

• Clarity: Be specific.
• Context: Include reaction conditions or setup.
• Structure: State expected response format.
• Examples: Use few-shot prompts when helpful.

Section 10.2 - Quiz Questions

1) Factual Questions

Question 1)

In a well‑structured prompt, the portion “SMILES string: CCO” is best labelled as:
A. General Instruction
B. Input Data
C. Relevant Domain
D. Few‑shot Example

Question 2)

Few‑shot examples are included primarily to:
A. Reduce token count
B. Demonstrate the output format and reasoning style the model should mimic
C. Hide the system instruction
D. Increase randomness

2) Comprehension / Application Questions

Question 3

You write the single‑line prompt: “Predict the major product.”
Your LLM responds with a 700‑word essay and no product structure. Which prompt‑engineering fix is most direct?
A. Add more temperature detail.
B. Specify “Output only a SMILES string, no commentary.”
C. Increase the temperature parameter to 2.
D. Split the prompt into two requests.

10.3 Usage of LLM APIs

Chemists can utilize APIs from OpenAI, Google, and others to:

• Embed intelligent text-based queries in chemical software or LIMS.
• Automate literature reviews: find publications, extract data, summarize research.
• Build dashboards for interactive chemical data analysis with AI insights.

Setting Up OpenAI API (Google Colab)

1. Open a notebook at https://colab.research.google.com
2. Install packages:

   !pip install openai 
   !pip install pandas
   !pip install numpy
   !pip install kagglehub
   

3. Authenticate:

   import os, openai
   os.environ["OPENAI_API_KEY"] = "YOUR_OPENAI_API_KEY"
   openai.api_key = os.getenv("OPENAI_API_KEY")
   

4. Load data:

   import kagglehub, pandas as pd
   path = kagglehub.dataset_download("priyanagda/bbbp-smiles")
   df = pd.read_csv(f"{path}/BBBP.csv")
   

   Or, if not on Kaggle:

   import requests, os, pandas as pd
   url = "URL_TO_FILE"
   os.makedirs("data", exist_ok=True)
   file_path = "data/file.csv"
   with open(file_path, "wb") as f:
   f.write(requests.get(url).content)
   df = pd.read_csv(file_path)
   

5. Make an API call:

   from openai import OpenAI
   client = OpenAI()
   sample_data = df.to_csv(index=False)
   prompt = f"YOUR QUESTION HERE. Reference data:\n{sample_data}"
   completion = client.chat.completions.create(
      model="gpt-4o-mini",
      messages=[
         {"role": "system", "content": "You are an expert chemist. Respond in Markdown."},
         {"role": "user",   "content": prompt}
      ],
      temperature=1,
      max_tokens=1000,
      top_p=1
   )
   

6. Evaluate:
   Call this function with two arguments, the model’s predicted answer and the ground-truth solution, to compute the model’s accuracy.

   def similarity_score(expected, answer):
   return int(expected.lower() == answer.lower())
   

Section 10.3 – Quiz Questions

1) Factual Questions

Question 1

In the Colab setup, the purpose of

os.environ["OPENAI_API_KEY"] = "YOUR_OPENAI_API_KEY"

is to:
A. Download a dataset from Kaggle
B. Make the key available to the openai library inside the session’s environment
C. Encrypt your code cell
D. Select the GPT‑4 model

Question 2

Which argument in client.chat.completions.create() mainly controls result length?
A. temperature
B. top_p
C. max_tokens
D. model

2) Comprehension / Application Questions

Question 3

You call the Chat Completion API twice with the exact same prompt but get different answers each time.
Which settings will make the response identical on every call?

A. temperature = 1.0, top_p = 1.0
B. temperature = 0.0, top_p = 1.0
C. temperature = 0.0, top_p = 0.0
D. temperature = 0.5, top_p = 0.5

10.4 Interactive Programming

Interactive programming means writing code incrementally with immediate feedback, typically in Jupyter Notebooks or Colab.

Colab Setup

1. Add secret:

• Sidebar → Secrets → Add new:
• Name: OPENAI_API_KEY
• Value: your key
• Enable notebook access

Prepare Library

2. Install libs:

   !pip install -q openai
   !pip install -q pandas

Coding

3. Helper function:

   import os
   import pandas as pd
   from google.colab import userdata
   from openai import OpenAI

   os.environ["OPENAI_API_KEY"] = userdata.get("OPENAI_API_KEY")
   client = OpenAI()

   def gpt_query(prompt, file_path=None):
      data_info = ""
      if file_path and file_path.endswith(".csv"):
         df = pd.read_csv(file_path)
         data_info = f"\nDataset sample:\n{df.head()}\n"
      completion = client.chat.completions.create(
         model="gpt-4o-mini",
         messages=[
            {"role": "system", "content": "You are a helpful assistant. Respond in Markdown."},
            {"role": "user",   "content": prompt + data_info}
         ],
         temperature=1,
         max_tokens=1000,
         top_p=1
      )
      return completion.choices[0].message.content.strip()

4. Run query:

   from IPython.display import Markdown, display
   output = gpt_query(prompt=input("Enter a prompt: "))
   display(Markdown(output))

Section 10.4 – Quiz Questions

1) Factual Questions

Question 1

Storing the API key in Colab “Secrets” is safer than typing it in plain text because:
A. It shortens execution time.
B. The key is hidden in the notebook UI and not saved to revision history.
C. It gives you free credits.
D. The model runs locally.

2) Comprehension / Application Questions

Question 2

The helper gpt_query() function optionally prints df.head() when given a .csv path.
Why is showing the head helpful before the API call?
A. To check column names and detect obvious parsing errors.
B. To compute cosine similarity.
C. To save network bandwidth.
D. To anonymise the dataset.