diff options
Diffstat (limited to 'notebooks/kg_rag_based_gpt_prompts.ipynb')
| -rw-r--r-- | notebooks/kg_rag_based_gpt_prompts.ipynb | 723 |
1 files changed, 723 insertions, 0 deletions
diff --git a/notebooks/kg_rag_based_gpt_prompts.ipynb b/notebooks/kg_rag_based_gpt_prompts.ipynb new file mode 100644 index 0000000..87dc281 --- /dev/null +++ b/notebooks/kg_rag_based_gpt_prompts.ipynb @@ -0,0 +1,723 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": 1, + "id": "3d3dca32-b77f-471d-b834-20ac795f9f17", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "os.chdir('..')\n" + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "id": "9da344d2-8e45-4574-aa19-4ad76c566101", + "metadata": {}, + "outputs": [], + "source": [ + "from IPython.display import clear_output\n", + "from kg_rag.utility import *\n", + "\n", + "clear_output()\n" + ] + }, + { + "cell_type": "code", + "execution_count": 3, + "id": "b44bf274-41d1-4153-a65e-bfb9b90ebcc6", + "metadata": {}, + "outputs": [], + "source": [ + "\n", + "def generate_response(question, llm, kg_rag_flag, evidence_flag=False, temperature=0):\n", + " CHAT_MODEL_ID = llm\n", + " CHAT_DEPLOYMENT_ID = llm\n", + " \n", + " if kg_rag_flag:\n", + " SYSTEM_PROMPT = system_prompts[\"KG_RAG_BASED_TEXT_GENERATION\"]\n", + " CONTEXT_VOLUME = int(config_data[\"CONTEXT_VOLUME\"])\n", + " QUESTION_VS_CONTEXT_SIMILARITY_PERCENTILE_THRESHOLD = float(config_data[\"QUESTION_VS_CONTEXT_SIMILARITY_PERCENTILE_THRESHOLD\"])\n", + " QUESTION_VS_CONTEXT_MINIMUM_SIMILARITY = float(config_data[\"QUESTION_VS_CONTEXT_MINIMUM_SIMILARITY\"])\n", + " VECTOR_DB_PATH = config_data[\"VECTOR_DB_PATH\"]\n", + " NODE_CONTEXT_PATH = config_data[\"NODE_CONTEXT_PATH\"]\n", + " SENTENCE_EMBEDDING_MODEL_FOR_NODE_RETRIEVAL = config_data[\"SENTENCE_EMBEDDING_MODEL_FOR_NODE_RETRIEVAL\"]\n", + " SENTENCE_EMBEDDING_MODEL_FOR_CONTEXT_RETRIEVAL = config_data[\"SENTENCE_EMBEDDING_MODEL_FOR_CONTEXT_RETRIEVAL\"]\n", + " vectorstore = load_chroma(VECTOR_DB_PATH, SENTENCE_EMBEDDING_MODEL_FOR_NODE_RETRIEVAL)\n", + " embedding_function_for_context_retrieval = load_sentence_transformer(SENTENCE_EMBEDDING_MODEL_FOR_CONTEXT_RETRIEVAL)\n", + " node_context_df = pd.read_csv(NODE_CONTEXT_PATH)\n", + " context = retrieve_context(question, vectorstore, embedding_function_for_context_retrieval, node_context_df, CONTEXT_VOLUME, QUESTION_VS_CONTEXT_SIMILARITY_PERCENTILE_THRESHOLD, QUESTION_VS_CONTEXT_MINIMUM_SIMILARITY, evidence_flag)\n", + " enriched_prompt = \"Context: \"+ context + \"\\n\" + \"Question: \" + question\n", + " question = enriched_prompt\n", + " else:\n", + " SYSTEM_PROMPT = system_prompts[\"PROMPT_BASED_TEXT_GENERATION\"]\n", + " \n", + " output = get_GPT_response(question, SYSTEM_PROMPT, CHAT_MODEL_ID, CHAT_DEPLOYMENT_ID, temperature=temperature)\n", + " stream_out(output)\n" + ] + }, + { + "cell_type": "code", + "execution_count": 4, + "id": "33c0771d-e6be-406b-9b17-51f6377bcb6a", + "metadata": {}, + "outputs": [], + "source": [ + "\n", + "LLM_TO_USE = 'gpt-4'\n", + "TEMPERATURE = config_data[\"LLM_TEMPERATURE\"]\n" + ] + }, + { + "cell_type": "markdown", + "id": "aab5d68e-ed47-4e36-986f-8842287d8ab6", + "metadata": {}, + "source": [ + "## Question 1:" + ] + }, + { + "cell_type": "code", + "execution_count": 5, + "id": "bbbdb428-6d01-43f2-9e58-b919e7a68736", + "metadata": {}, + "outputs": [], + "source": [ + "question = 'Are there any latest drugs used for weight management in patients with Bardet-Biedl Syndrome?'\n" + ] + }, + { + "cell_type": "markdown", + "id": "19ffb33b-bce6-4cb1-8f86-76080fcf0e40", + "metadata": {}, + "source": [ + "### With KG-RAG" + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "id": "02374d3c-5711-4704-8b37-7eda0965c9b0", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Yes, the compound Setmelanotide is used to treat Bardet-Biedl syndrome. It is currently in phase 3 of clinical trials according to the sources ChEMBL and DrugCentral. However, it is advised to seek guidance from a healthcare professional for the most current and personalized treatment options. [Provenance: ChEMBL, DrugCentral]\n", + "\n" + ] + } + ], + "source": [ + "KG_RAG_FLAG = True\n", + "EDGE_EVIDENCE_FLAG = True #Used only when KG_RAG_FLAG=True\n", + "\n", + "generate_response(question, LLM_TO_USE, KG_RAG_FLAG, evidence_flag=EDGE_EVIDENCE_FLAG, temperature=TEMPERATURE)\n" + ] + }, + { + "cell_type": "markdown", + "id": "17be2ecd-40f1-44b5-8f79-9358264a3703", + "metadata": {}, + "source": [ + "### Without KG-RAG" + ] + }, + { + "cell_type": "code", + "execution_count": 7, + "id": "06bf55f2-eccf-4c26-b65a-0ed2ed30e689", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "As of my knowledge up to date, there are no specific drugs designed for weight management in patients with Bardet-Biedl Syndrome. The treatment generally involves managing the symptoms and complications. However, any new developments would be best advised by a healthcare professional.\n", + "\n" + ] + } + ], + "source": [ + "KG_RAG_FLAG = False\n", + "\n", + "generate_response(question, LLM_TO_USE, KG_RAG_FLAG, temperature=TEMPERATURE)\n" + ] + }, + { + "cell_type": "markdown", + "id": "afa24795-71e1-4f1a-ac99-b8d106513f76", + "metadata": {}, + "source": [ + "## Question 2:" + ] + }, + { + "cell_type": "code", + "execution_count": 8, + "id": "eca290bb-fe30-4bb2-ab36-405c2151bccb", + "metadata": {}, + "outputs": [], + "source": [ + "question = 'Is it PNPLA3 or HLA-B that has a significant association with the disease liver benign neoplasm?'\n" + ] + }, + { + "cell_type": "markdown", + "id": "0cd8e67f-f769-4130-8e7d-52e0524c66df", + "metadata": {}, + "source": [ + "### With KG-RAG" + ] + }, + { + "cell_type": "code", + "execution_count": 9, + "id": "03863049-4ea5-4733-aeb4-0be3f09d6b5f", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "The gene PNPLA3 has a more significant association with the disease liver benign neoplasm, as indicated by the lower p-value of 4e-14 compared to HLA-B's p-value of 2e-08. The provenance of this information is the GWAS Catalog.\n", + "\n" + ] + } + ], + "source": [ + "KG_RAG_FLAG = True\n", + "EDGE_EVIDENCE_FLAG = True #Used only when KG_RAG_FLAG=True\n", + "\n", + "generate_response(question, LLM_TO_USE, KG_RAG_FLAG, evidence_flag=EDGE_EVIDENCE_FLAG, temperature=TEMPERATURE)\n" + ] + }, + { + "cell_type": "markdown", + "id": "30121a5b-3956-41bd-ac4a-62e32744ba99", + "metadata": {}, + "source": [ + "### Without KG-RAG" + ] + }, + { + "cell_type": "code", + "execution_count": 10, + "id": "8614706d-90d9-49e8-9481-30628505ba2e", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "It is PNPLA3 that has a significant association with the disease liver benign neoplasm.\n", + "\n" + ] + } + ], + "source": [ + "KG_RAG_FLAG = False\n", + "\n", + "generate_response(question, LLM_TO_USE, KG_RAG_FLAG, temperature=TEMPERATURE)\n" + ] + }, + { + "cell_type": "markdown", + "id": "8087eb26-e144-48ee-8630-09c73eb35a37", + "metadata": {}, + "source": [ + "## Question 3" + ] + }, + { + "cell_type": "code", + "execution_count": 11, + "id": "4ed3bb71-12c8-40bf-b361-f9cd227c142d", + "metadata": {}, + "outputs": [], + "source": [ + "question = \"Is Parkinson's disease associated with PINK1 gene?\"\n" + ] + }, + { + "cell_type": "code", + "execution_count": 12, + "id": "0d06bb9e-ed6e-4f1b-ac2e-65cf219eeb12", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Yes, Parkinson's disease is associated with the PINK1 gene. This association is reported in the DISEASES database - https://diseases.jensenlab.org.\n", + "\n" + ] + } + ], + "source": [ + "KG_RAG_FLAG = True\n", + "EDGE_EVIDENCE_FLAG = True \n", + "\n", + "generate_response(question, LLM_TO_USE, KG_RAG_FLAG, evidence_flag=EDGE_EVIDENCE_FLAG, temperature=TEMPERATURE)\n" + ] + }, + { + "cell_type": "markdown", + "id": "b17ed61e-9de1-4e2b-ad04-5c41b1167e5b", + "metadata": {}, + "source": [ + "## Question 3- perturbed (entities in smaller case)" + ] + }, + { + "cell_type": "code", + "execution_count": 13, + "id": "ae0cccdb-8689-46b0-a1cb-808ec8f997e0", + "metadata": {}, + "outputs": [], + "source": [ + "question = \"Is parkinson's disease associated with pink1 gene?\"\n" + ] + }, + { + "cell_type": "code", + "execution_count": 14, + "id": "1961ca49-f48a-46e3-bfc3-ced02fb1a2c2", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Yes, Parkinson's disease is associated with the PINK1 gene. This association is reported in the DISEASES database - https://diseases.jensenlab.org.\n", + "\n" + ] + } + ], + "source": [ + "KG_RAG_FLAG = True\n", + "EDGE_EVIDENCE_FLAG = True \n", + "\n", + "generate_response(question, LLM_TO_USE, KG_RAG_FLAG, evidence_flag=EDGE_EVIDENCE_FLAG, temperature=TEMPERATURE)\n" + ] + }, + { + "cell_type": "markdown", + "id": "b7f6ee15-2e3a-4dea-bfa6-35401b6b67da", + "metadata": {}, + "source": [ + "## Question 4:" + ] + }, + { + "cell_type": "code", + "execution_count": 15, + "id": "f5068a03-143d-4c29-87bf-bba1082408bd", + "metadata": {}, + "outputs": [], + "source": [ + "question = \"What are some protein markers associated with thoracic aortic aneurysm?\"\n" + ] + }, + { + "cell_type": "markdown", + "id": "056a4de0-c0fb-48ee-93e3-7742584362c9", + "metadata": {}, + "source": [ + "### With KG-RAG" + ] + }, + { + "cell_type": "code", + "execution_count": 16, + "id": "6edd6bc4-e03b-473e-996e-6705f83ee2fa", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "The protein markers associated with thoracic aortic aneurysm include Chondroitin sulfate proteoglycan 4 (CSPG4), Matrix Gla protein (MGP), Interleukin-2 receptor subunit alpha (IL2RA), Interleukin-1 beta (IL1B), Myosin-10 (MYH10), Tropomyosin alpha-4 chain (TPM4), Tyrosine-protein kinase Mer (MERTK), and Stabilin-1 (STAB1). The provenance of these associations is the Cell Taxonomy database.\n", + "\n" + ] + } + ], + "source": [ + "KG_RAG_FLAG = True\n", + "EDGE_EVIDENCE_FLAG = True \n", + "\n", + "generate_response(question, LLM_TO_USE, KG_RAG_FLAG, evidence_flag=EDGE_EVIDENCE_FLAG, temperature=TEMPERATURE)\n" + ] + }, + { + "cell_type": "markdown", + "id": "e95cd283-ebf1-4394-bf40-e62fc96505eb", + "metadata": {}, + "source": [ + "### Without KG-RAG" + ] + }, + { + "cell_type": "code", + "execution_count": 17, + "id": "f4c9d96b-c78f-44d4-9a46-1254f4856324", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Some protein markers associated with thoracic aortic aneurysm include Matrix metalloproteinases (MMPs), C-reactive protein (CRP), and fibrillin-1.\n", + "\n" + ] + } + ], + "source": [ + "KG_RAG_FLAG = False\n", + "\n", + "generate_response(question, LLM_TO_USE, KG_RAG_FLAG, temperature=TEMPERATURE)\n" + ] + }, + { + "cell_type": "markdown", + "id": "84ecf6b4-5732-453a-94c3-a524c4572b81", + "metadata": {}, + "source": [ + "## Question 5:" + ] + }, + { + "cell_type": "code", + "execution_count": 18, + "id": "0a049b56-cc9c-4777-8351-265839145690", + "metadata": {}, + "outputs": [], + "source": [ + "question = \"Are there any protein markers that show increased activity in adenocarcinoma?\"\n" + ] + }, + { + "cell_type": "markdown", + "id": "a9a82988-b1e6-441b-bafa-4b148e4054ba", + "metadata": {}, + "source": [ + "### With KG-RAG" + ] + }, + { + "cell_type": "code", + "execution_count": 19, + "id": "6cd1217e-b504-4006-a099-87bc5361d85f", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Yes, there are several protein markers that show increased activity in adenocarcinoma. These include Keratin, type II cytoskeletal 7 (Cytokeratin-7) (CK-7) (Keratin-7) (K7) (Sarcolectin) (Type-II keratin Kb7), Anterior gradient protein 2 homolog (AG-2) (hAG-2) (HPC8) (Secreted cement gland protein XAG-2 homolog), Guanine deaminase (Guanase) (Guanine aminase) (3.5.4.3) (Guanine aminohydrolase) (GAH) (p51-nedasin), and Graves disease carrier protein (GDC) (Graves disease autoantigen) (GDA) (Mitochondrial solute carrier protein homolog) (Solute carrier family 25 member 16). The provenance of these associations is the Cell Taxonomy.\n", + "\n" + ] + } + ], + "source": [ + "KG_RAG_FLAG = True\n", + "EDGE_EVIDENCE_FLAG = True \n", + "\n", + "generate_response(question, LLM_TO_USE, KG_RAG_FLAG, evidence_flag=EDGE_EVIDENCE_FLAG, temperature=TEMPERATURE)\n" + ] + }, + { + "cell_type": "markdown", + "id": "7bcae3c8-130c-4121-a119-9b9bb79a21ee", + "metadata": {}, + "source": [ + "### Without KG-RAG" + ] + }, + { + "cell_type": "code", + "execution_count": 20, + "id": "96d8e70c-4258-4679-a5fa-1bc798eabd59", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Yes, certain protein markers like carcinoembryonic antigen (CEA), CA 19-9, and cytokeratins can show increased activity in adenocarcinoma.\n", + "\n" + ] + } + ], + "source": [ + "KG_RAG_FLAG = False\n", + "\n", + "generate_response(question, LLM_TO_USE, KG_RAG_FLAG, temperature=TEMPERATURE)\n" + ] + }, + { + "cell_type": "markdown", + "id": "03697e1c-6868-4c56-8876-1333bc493f44", + "metadata": {}, + "source": [ + "## Question 6:" + ] + }, + { + "cell_type": "code", + "execution_count": 21, + "id": "189ef273-bf3a-415f-8198-0e7fc4a1cac0", + "metadata": {}, + "outputs": [], + "source": [ + "question = \"Do you know if ruxolitinib is approved as a pharmacologic treatment for vitiligo?\"\n" + ] + }, + { + "cell_type": "markdown", + "id": "fffd4c16-08fa-436f-b692-d3fe9eda305e", + "metadata": {}, + "source": [ + "### With KG-RAG" + ] + }, + { + "cell_type": "code", + "execution_count": 22, + "id": "8e520c41-c5d3-4b38-a744-f954c106847a", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Yes, Ruxolitinib is associated with the treatment of vitiligo. This association is supported by data from ChEMBL and DrugCentral databases. However, it is always recommended to seek guidance from a healthcare professional for treatment options. (Provenance: ChEMBL, DrugCentral)\n", + "\n" + ] + } + ], + "source": [ + "KG_RAG_FLAG = True\n", + "EDGE_EVIDENCE_FLAG = True \n", + "\n", + "generate_response(question, LLM_TO_USE, KG_RAG_FLAG, evidence_flag=EDGE_EVIDENCE_FLAG, temperature=TEMPERATURE)\n" + ] + }, + { + "cell_type": "markdown", + "id": "6b1c1e0c-2a6a-48a0-99f1-4daabb2f1f61", + "metadata": {}, + "source": [ + "### Without KG-RAG" + ] + }, + { + "cell_type": "code", + "execution_count": 23, + "id": "91e4eabc-2027-4bb6-b75f-0064d2a52369", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "As of my knowledge up to date, ruxolitinib is not officially approved for the treatment of vitiligo. However, some clinical trials have shown promising results. Always consult with a healthcare provider for treatment options.\n", + "\n" + ] + } + ], + "source": [ + "KG_RAG_FLAG = False\n", + "\n", + "generate_response(question, LLM_TO_USE, KG_RAG_FLAG, temperature=TEMPERATURE)\n" + ] + }, + { + "cell_type": "markdown", + "id": "be6ca1c2-f118-41cc-93a7-5e8db134193f", + "metadata": {}, + "source": [ + "## Question 7:" + ] + }, + { + "cell_type": "code", + "execution_count": 24, + "id": "36a704be-f9d6-4968-82ec-09c4d0efe3a2", + "metadata": {}, + "outputs": [], + "source": [ + "question = \"Are there any biomarkers that show increased profile in hydrocephalus?\"\n" + ] + }, + { + "cell_type": "markdown", + "id": "cf579442-9b23-4160-be25-e4a9dd88999f", + "metadata": {}, + "source": [ + "### With KG-RAG" + ] + }, + { + "cell_type": "code", + "execution_count": 25, + "id": "a9f242a8-4f31-4f58-926c-d48f3f95f584", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Yes, the Transmembrane protein 119 (TMEM119) and P2Y purinoceptor 12 (P2RY12) show an increased profile in hydrocephalus. The provenance of this information is the Cell Taxonomy database.\n", + "\n" + ] + } + ], + "source": [ + "KG_RAG_FLAG = True\n", + "EDGE_EVIDENCE_FLAG = True \n", + "\n", + "generate_response(question, LLM_TO_USE, KG_RAG_FLAG, evidence_flag=EDGE_EVIDENCE_FLAG, temperature=TEMPERATURE)\n" + ] + }, + { + "cell_type": "markdown", + "id": "e0904315-0f47-4422-aa5c-063a3ccbc919", + "metadata": {}, + "source": [ + "### Without KG-RAG" + ] + }, + { + "cell_type": "code", + "execution_count": 26, + "id": "1f1ab642-9385-4b04-83cf-6ace3eee6653", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Yes, certain biomarkers such as L1CAM, S100B, GFAP, and NSE have shown increased profiles in hydrocephalus.\n", + "\n" + ] + } + ], + "source": [ + "KG_RAG_FLAG = False\n", + "\n", + "generate_response(question, LLM_TO_USE, KG_RAG_FLAG, temperature=TEMPERATURE)\n" + ] + }, + { + "cell_type": "markdown", + "id": "1fd3811b-546f-4ec3-8207-f9323e372744", + "metadata": {}, + "source": [ + "## Question 8:" + ] + }, + { + "cell_type": "code", + "execution_count": 27, + "id": "2a1c9337-fd39-45b0-b12a-6de9b5971b9e", + "metadata": {}, + "outputs": [], + "source": [ + "\n", + "question = 'Does drug dependence have any genetic factors? Do you have any statistical evidence from trustworthy sources for this?'\n" + ] + }, + { + "cell_type": "markdown", + "id": "93f43325-fd99-4b43-a2b3-1314c26c7d8e", + "metadata": {}, + "source": [ + "### With KG-RAG" + ] + }, + { + "cell_type": "code", + "execution_count": 28, + "id": "e6852cb3-8bf9-408b-ab65-492b75c690ed", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Yes, drug dependence does have genetic factors. The genes KAT2B and SLC25A16 have been associated with drug dependence. This information is backed by statistical evidence from the GWAS Catalog, with p-values of 4e-10 and 1e-09 respectively.\n", + "\n" + ] + } + ], + "source": [ + "KG_RAG_FLAG = True\n", + "EDGE_EVIDENCE_FLAG = True \n", + "\n", + "generate_response(question, LLM_TO_USE, KG_RAG_FLAG, evidence_flag=EDGE_EVIDENCE_FLAG, temperature=TEMPERATURE)\n" + ] + }, + { + "cell_type": "markdown", + "id": "796144e2-338d-479c-ba3b-34b0c23a209b", + "metadata": {}, + "source": [ + "### Without KG-RAG" + ] + }, + { + "cell_type": "code", + "execution_count": 29, + "id": "b8079bc6-d309-4c88-9440-376aa43d972e", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Yes, drug dependence does have genetic factors. According to the National Institute on Drug Abuse, genetics account for about 40-60% of a person's vulnerability to drug addiction.\n", + "\n" + ] + } + ], + "source": [ + "KG_RAG_FLAG = False\n", + "\n", + "generate_response(question, LLM_TO_USE, KG_RAG_FLAG, temperature=TEMPERATURE)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "16e1ae6d-d0e0-4b42-a8fe-517033fb9960", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.10.9" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} |