{"id":79776,"topic":"ai","source":"WVIR","title":"UVA School of Medicine researchers find promise, pitfalls in AI models for biomedical research - WVIR","url":"https://www.29news.com/2026/09/10/uva-school-medicine-researchers-find-promise-pitfalls-ai-models-biomedical-research/","url_hash":"3e7b941798e39281027f50855d8e7482f08c108d","author":"","summary":"<a href=\"https://news.google.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?oc=5\" target=\"_blank\">UVA School of Medicine researchers find promise, pitfalls in AI models for biomedical research</a>&nbsp;&nbsp;<font color=\"#6f6f6f\">WVIR</font>","content":"CHARLOTTESVILLE, Va. (WVIR) - A new University of Virginia study tested popular artificial intelligence tools, including ChatGPT, Gemini and Claude, by asking them to explain how cells in the body communicate.\nResearchers focused on heart cells and found the AI could correctly identify up to 65% of known cell reactions.\nWhen it came to predicting how those cells respond to disease or new drugs, the AI got it right as little as 6% of the time.\nUVA says that gap matters because getting it wrong in a lab can mean millions of dollars lost, and false hope for patients.\nJeff Saucerman, PhD, of UVA’s Department of Biomedical Engineering, said the technology is improving.\n“It’s pretty good already at knowing the individual pieces of cells, but it’s not very good at piecing them together, and that’s what’s needed to predict more substantially into the future what new drugs would do,” Saucerman said.\nSaucerman said testing AI models at multiple levels is necessary before relying on their predictions.\n“It was absolutely critical that they get tested at many different levels by different approaches,” Saucerman said. “Just because a model makes a prediction doesn’t mean we should trust it.”\nSaucerman says the technology is improving fast, but for now, researchers say human scientists still need to have the final say.\nDo you have a story idea? Send us your news tip here.\nCopyright 2026 WVIR. All rights reserved.","image_url":"https://gray-wvir-prod.gtv-cdn.com/resizer/v2/XVRNFHCXJJD4LGHKHRTSACJUVU.png?auth=7a1bca51bbc53cbb375ef1dea8eef178e9336c11e3a48cfdcfe371887b35b203&width=1200&height=600&smart=true","lang":"en","published_at":"2026-09-10T21:52:00+00:00","fetched_at":"2026-09-11T00:15:03+00:00","status":"read","starred":0,"extract_state":"ok","summary_auto":"(WVIR) - A new University of Virginia study tested popular artificial intelligence tools, including ChatGPT, Gemini and Claude, by asking them to explain how cells in the body communicate. Researchers focused on heart cells and found the AI could correctly identify up to 65% of known cell reactions.","cluster_id":null,"extract_retries":0,"extract_error":null,"contract_version":"news_item.v1","format_contract_version":"news_item_formats.v1","dedup_url":"https://www.29news.com/2026/09/10/uva-school-medicine-researchers-find-promise-pitfalls-ai-models-biomedical-research/","quality_profile":{"profile_version":"extraction_quality.v2","bucket":"high","confidence":0.9,"failure_kind":"none","retryable":false,"retry_after_attempts":0,"reason":"High confidence: full text extraction produced 1421 characters.","operator_guidance":{"severity":"ok","recommended_action":"trust_full_text","next_step":"Use the extracted full text as the primary article source.","operator_label":"Ready","can_retry":false,"can_use_summary":false,"diagnostics_required":false},"content_depth":{"contract_version":"content_depth.v1","category":"full_text","label":"Full text","has_full_text":true,"has_summary":true,"content_length":1421,"summary_length":300,"usable_text_length":1421,"source_field":"content"},"legacy_collapsed":false,"signals":{"extract_state":"ok","extract_error":null,"extract_retries":0,"content_length":1421,"summary_length":300}},"news_item":{"id":79776,"canonical_url":"https://www.29news.com/2026/09/10/uva-school-medicine-researchers-find-promise-pitfalls-ai-models-biomedical-research/","source_url":"https://www.29news.com/2026/09/10/uva-school-medicine-researchers-find-promise-pitfalls-ai-models-biomedical-research/","title":"UVA School of Medicine researchers find promise, pitfalls in AI models for biomedical research - WVIR","source_name":"WVIR","author":null,"published_at":"2026-09-10T21:52:00+00:00","locale":"en","topic":"ai","tags":[],"rss_summary":"<a href=\"https://news.google.com/rss/articles/CBMiugFBVV95cUxPSHZlVHgydk9NQm4yaEFlaENLUFB1ZWpvSHFxTHpueGp2bkxUODBQcHRCUXM4c0JUMW03UzRqcHZ2V2UzamtGWldCN3ZBcnQ0Y2Z2N3dOdTQzbmtQcjNCNk9oU1Blc1dnUFZHa0VtVmtGcjUwN2Nkc2ZUbF9yVTgxVEd3U05fNFVnMFFySzc3ZEVoUjFJMldfMldqVjdlSVY5QS0wYjE2XzdMWEhiMlZINGRwVk9uSjNEYVHSAc4BQVVfeXFMTl9IVzV2WXZaVkxvQjBsd1RMU0xYaUh3UFFoeEhINnF2endBWkNBaWhzLU4yTjYzNUxCblY1WVdYb1JaZWdQX2xMSmJrcmFMM3dJeURCazRjWnNkaFVvTnl3QVBBVnRuaHFqYXUtUkI5V2I2UzJoOHAwX0I5QURVMl92UzI2TERsZHJHaUxNRzlwZ2hkT3V3MmFuMWtiY0twcG56SXhocDQtVjk5bzRQZlhKYXFRSkdaYXhNak1SNVlpWElQcmdYb0FVMzNSNGc?oc=5\" target=\"_blank\">UVA School of Medicine researchers find promise, pitfalls in AI models for biomedical research</a>&nbsp;&nbsp;<font color=\"#6f6f6f\">WVIR</font>","full_text":"CHARLOTTESVILLE, Va. (WVIR) - A new University of Virginia study tested popular artificial intelligence tools, including ChatGPT, Gemini and Claude, by asking them to explain how cells in the body communicate.\nResearchers focused on heart cells and found the AI could correctly identify up to 65% of known cell reactions.\nWhen it came to predicting how those cells respond to disease or new drugs, the AI got it right as little as 6% of the time.\nUVA says that gap matters because getting it wrong in a lab can mean millions of dollars lost, and false hope for patients.\nJeff Saucerman, PhD, of UVA’s Department of Biomedical Engineering, said the technology is improving.\n“It’s pretty good already at knowing the individual pieces of cells, but it’s not very good at piecing them together, and that’s what’s needed to predict more substantially into the future what new drugs would do,” Saucerman said.\nSaucerman said testing AI models at multiple levels is necessary before relying on their predictions.\n“It was absolutely critical that they get tested at many different levels by different approaches,” Saucerman said. “Just because a model makes a prediction doesn’t mean we should trust it.”\nSaucerman says the technology is improving fast, but for now, researchers say human scientists still need to have the final say.\nDo you have a story idea? Send us your news tip here.\nCopyright 2026 WVIR. All rights reserved.","excerpt":"(WVIR) - A new University of Virginia study tested popular artificial intelligence tools, including ChatGPT, Gemini and Claude, by asking them to explain how cells in the body communicate. Researchers focused on heart cells and found the AI could correctly identify up to 65% of known cell reactions.","extraction":{"state":"ok","confidence":0.9,"error":null,"explanation":"High confidence: full text extraction produced 1421 characters.","diagnostics_url":"/api/diagnose?url=https%3A//www.29news.com/2026/09/10/uva-school-medicine-researchers-find-promise-pitfalls-ai-models-biomedical-research/","quality_profile":{"profile_version":"extraction_quality.v2","bucket":"high","confidence":0.9,"failure_kind":"none","retryable":false,"retry_after_attempts":0,"reason":"High confidence: full text extraction produced 1421 characters.","operator_guidance":{"severity":"ok","recommended_action":"trust_full_text","next_step":"Use the extracted full text as the primary article source.","operator_label":"Ready","can_retry":false,"can_use_summary":false,"diagnostics_required":false},"content_depth":{"contract_version":"content_depth.v1","category":"full_text","label":"Full text","has_full_text":true,"has_summary":true,"content_length":1421,"summary_length":300,"usable_text_length":1421,"source_field":"content"},"legacy_collapsed":false,"signals":{"extract_state":"ok","extract_error":null,"extract_retries":0,"content_length":1421,"summary_length":300}}},"display_formats":["compact","card","full","digest_section","json"]},"daily_stack_record":{"title":"UVA School of Medicine researchers find promise, pitfalls in AI models for biomedical research - WVIR","url":"https://www.29news.com/2026/09/10/uva-school-medicine-researchers-find-promise-pitfalls-ai-models-biomedical-research/","summary":"(WVIR) - A new University of Virginia study tested popular artificial intelligence tools, including ChatGPT, Gemini and Claude, by asking them to explain how cells in the body communicate. Researchers focused on heart cells and found the AI could correctly identify up to 65% of known cell reactions.","source":"WVIR","date":"2026-09-10T21:52:00+00:00","content":"CHARLOTTESVILLE, Va. (WVIR) - A new University of Virginia study tested popular artificial intelligence tools, including ChatGPT, Gemini and Claude, by asking them to explain how cells in the body communicate.\nResearchers focused on heart cells and found the AI could correctly identify up to 65% of known cell reactions.\nWhen it came to predicting how those cells respond to disease or new drugs, the AI got it right as little as 6% of the time.\nUVA says that gap matters because getting it wrong in a lab can mean millions of dollars lost, and false hope for patients.\nJeff Saucerman, PhD, of UVA’s Department of Biomedical Engineering, said the technology is improving.\n“It’s pretty good already at knowing the individual pieces of cells, but it’s not very good at piecing them together, and that’s what’s needed to predict more substantially into the future what new drugs would do,” Saucerman said.\nSaucerman said testing AI models at multiple levels is necessary before relying on their predictions.\n“It was absolutely critical that they get tested at many different levels by different approaches,” Saucerman said. “Just because a model makes a prediction doesn’t mean we should trust it.”\nSaucerman says the technology is improving fast, but for now, researchers say human scientists still need to have the final say.\nDo you have a story idea? Send us your news tip here.\nCopyright 2026 WVIR. All rights reserved.","confidence":0.9,"diagnostics_url":"/api/diagnose?url=https%3A//www.29news.com/2026/09/10/uva-school-medicine-researchers-find-promise-pitfalls-ai-models-biomedical-research/","quality_bucket":"high","failure_kind":"none","retryable":false,"quality_reason":"High confidence: full text extraction produced 1421 characters.","quality_profile":{"profile_version":"extraction_quality.v2","bucket":"high","confidence":0.9,"failure_kind":"none","retryable":false,"retry_after_attempts":0,"reason":"High confidence: full text extraction produced 1421 characters.","operator_guidance":{"severity":"ok","recommended_action":"trust_full_text","next_step":"Use the extracted full text as the primary article source.","operator_label":"Ready","can_retry":false,"can_use_summary":false,"diagnostics_required":false},"content_depth":{"contract_version":"content_depth.v1","category":"full_text","label":"Full text","has_full_text":true,"has_summary":true,"content_length":1421,"summary_length":300,"usable_text_length":1421,"source_field":"content"},"legacy_collapsed":false,"signals":{"extract_state":"ok","extract_error":null,"extract_retries":0,"content_length":1421,"summary_length":300}},"tags":[]},"fallback_formats":["markdown","json","html"],"actions":{"read":"/item/79776","export_markdown":"/api/items/79776/export?format=markdown","export_json":"/api/items/79776/export?format=json","diagnose":"/api/diagnose?url=https%3A//www.29news.com/2026/09/10/uva-school-medicine-researchers-find-promise-pitfalls-ai-models-biomedical-research/"},"formats":{"full":{"id":79776,"title":"UVA School of Medicine researchers find promise, pitfalls in AI models for biomedical research - WVIR","url":"https://www.29news.com/2026/09/10/uva-school-medicine-researchers-find-promise-pitfalls-ai-models-biomedical-research/","source":"WVIR","author":null,"published_at":"2026-09-10T21:52:00+00:00","locale":"en","topic":"ai","tags":[],"excerpt":"(WVIR) - A new University of Virginia study tested popular artificial intelligence tools, including ChatGPT, Gemini and Claude, by asking them to explain how cells in the body communicate. Researchers focused on heart cells and found the AI could correctly identify up to 65% of known cell reactions.","full_text":"CHARLOTTESVILLE, Va. (WVIR) - A new University of Virginia study tested popular artificial intelligence tools, including ChatGPT, Gemini and Claude, by asking them to explain how cells in the body communicate.\nResearchers focused on heart cells and found the AI could correctly identify up to 65% of known cell reactions.\nWhen it came to predicting how those cells respond to disease or new drugs, the AI got it right as little as 6% of the time.\nUVA says that gap matters because getting it wrong in a lab can mean millions of dollars lost, and false hope for patients.\nJeff Saucerman, PhD, of UVA’s Department of Biomedical Engineering, said the technology is improving.\n“It’s pretty good already at knowing the individual pieces of cells, but it’s not very good at piecing them together, and that’s what’s needed to predict more substantially into the future what new drugs would do,” Saucerman said.\nSaucerman said testing AI models at multiple levels is necessary before relying on their predictions.\n“It was absolutely critical that they get tested at many different levels by different approaches,” Saucerman said. “Just because a model makes a prediction doesn’t mean we should trust it.”\nSaucerman says the technology is improving fast, but for now, researchers say human scientists still need to have the final say.\nDo you have a story idea? Send us your news tip here.\nCopyright 2026 WVIR. All rights reserved.","reading_time_min":1,"extraction":{"state":"ok","confidence":0.9,"error":null,"explanation":"High confidence: full text extraction produced 1421 characters.","diagnostics_url":"/api/diagnose?url=https%3A//www.29news.com/2026/09/10/uva-school-medicine-researchers-find-promise-pitfalls-ai-models-biomedical-research/","quality_profile":{"profile_version":"extraction_quality.v2","bucket":"high","confidence":0.9,"failure_kind":"none","retryable":false,"retry_after_attempts":0,"reason":"High confidence: full text extraction produced 1421 characters.","operator_guidance":{"severity":"ok","recommended_action":"trust_full_text","next_step":"Use the extracted full text as the primary article source.","operator_label":"Ready","can_retry":false,"can_use_summary":false,"diagnostics_required":false},"content_depth":{"contract_version":"content_depth.v1","category":"full_text","label":"Full text","has_full_text":true,"has_summary":true,"content_length":1421,"summary_length":300,"usable_text_length":1421,"source_field":"content"},"legacy_collapsed":false,"signals":{"extract_state":"ok","extract_error":null,"extract_retries":0,"content_length":1421,"summary_length":300}}},"quality_profile":{"profile_version":"extraction_quality.v2","bucket":"high","confidence":0.9,"failure_kind":"none","retryable":false,"retry_after_attempts":0,"reason":"High confidence: full text extraction produced 1421 characters.","operator_guidance":{"severity":"ok","recommended_action":"trust_full_text","next_step":"Use the extracted full text as the primary article source.","operator_label":"Ready","can_retry":false,"can_use_summary":false,"diagnostics_required":false},"content_depth":{"contract_version":"content_depth.v1","category":"full_text","label":"Full text","has_full_text":true,"has_summary":true,"content_length":1421,"summary_length":300,"usable_text_length":1421,"source_field":"content"},"legacy_collapsed":false,"signals":{"extract_state":"ok","extract_error":null,"extract_retries":0,"content_length":1421,"summary_length":300}},"actions":{"read":"/item/79776","export_markdown":"/api/items/79776/export?format=markdown","export_json":"/api/items/79776/export?format=json","diagnose":"/api/diagnose?url=https%3A//www.29news.com/2026/09/10/uva-school-medicine-researchers-find-promise-pitfalls-ai-models-biomedical-research/"}},"digest":{"id":79776,"title":"UVA School of Medicine researchers find promise, pitfalls in AI models for biomedical research - WVIR","url":"https://www.29news.com/2026/09/10/uva-school-medicine-researchers-find-promise-pitfalls-ai-models-biomedical-research/","source":"WVIR","topic":"ai","published_at":"2026-09-10T21:52:00+00:00","excerpt":"(WVIR) - A new University of Virginia study tested popular artificial intelligence tools, including ChatGPT, Gemini and Claude, by asking them to explain how cells in the body communicate. Researchers focused on heart cells and found the AI could correctly identify up to 65% of…","quality_bucket":"high","quality_reason":"High confidence: full text extraction produced 1421 characters.","reading_time_min":1,"cluster_id":null},"card":{"display_title":"UVA School of Medicine researchers find promise, pitfalls in AI models for biomedical research - WVIR","subtitle":"WVIR · 2026-09-10","summary":"(WVIR) - A new University of Virginia study tested popular artificial intelligence tools, including ChatGPT, Gemini and Claude, by asking them to explain how cells in the body communicate. Researchers focused on heart…","badges":["quality:high"],"links":{"read":"/item/79776","original":"https://www.29news.com/2026/09/10/uva-school-medicine-researchers-find-promise-pitfalls-ai-models-biomedical-research/","diagnose":"/api/diagnose?url=https%3A//www.29news.com/2026/09/10/uva-school-medicine-researchers-find-promise-pitfalls-ai-models-biomedical-research/"},"quality_warning":null},"export":{"title":"UVA School of Medicine researchers find promise, pitfalls in AI models for biomedical research - WVIR","url":"https://www.29news.com/2026/09/10/uva-school-medicine-researchers-find-promise-pitfalls-ai-models-biomedical-research/","summary":"(WVIR) - A new University of Virginia study tested popular artificial intelligence tools, including ChatGPT, Gemini and Claude, by asking them to explain how cells in the body communicate. Researchers focused on heart cells and found the AI could correctly identify up to 65% of known cell reactions.","source":"WVIR","date":"2026-09-10T21:52:00+00:00","content":"CHARLOTTESVILLE, Va. (WVIR) - A new University of Virginia study tested popular artificial intelligence tools, including ChatGPT, Gemini and Claude, by asking them to explain how cells in the body communicate.\nResearchers focused on heart cells and found the AI could correctly identify up to 65% of known cell reactions.\nWhen it came to predicting how those cells respond to disease or new drugs, the AI got it right as little as 6% of the time.\nUVA says that gap matters because getting it wrong in a lab can mean millions of dollars lost, and false hope for patients.\nJeff Saucerman, PhD, of UVA’s Department of Biomedical Engineering, said the technology is improving.\n“It’s pretty good already at knowing the individual pieces of cells, but it’s not very good at piecing them together, and that’s what’s needed to predict more substantially into the future what new drugs would do,” Saucerman said.\nSaucerman said testing AI models at multiple levels is necessary before relying on their predictions.\n“It was absolutely critical that they get tested at many different levels by different approaches,” Saucerman said. “Just because a model makes a prediction doesn’t mean we should trust it.”\nSaucerman says the technology is improving fast, but for now, researchers say human scientists still need to have the final say.\nDo you have a story idea? Send us your news tip here.\nCopyright 2026 WVIR. All rights reserved.","confidence":0.9,"diagnostics_url":"/api/diagnose?url=https%3A//www.29news.com/2026/09/10/uva-school-medicine-researchers-find-promise-pitfalls-ai-models-biomedical-research/","quality_bucket":"high","failure_kind":"none","retryable":false,"quality_reason":"High confidence: full text extraction produced 1421 characters.","quality_profile":{"profile_version":"extraction_quality.v2","bucket":"high","confidence":0.9,"failure_kind":"none","retryable":false,"retry_after_attempts":0,"reason":"High confidence: full text extraction produced 1421 characters.","operator_guidance":{"severity":"ok","recommended_action":"trust_full_text","next_step":"Use the extracted full text as the primary article source.","operator_label":"Ready","can_retry":false,"can_use_summary":false,"diagnostics_required":false},"content_depth":{"contract_version":"content_depth.v1","category":"full_text","label":"Full text","has_full_text":true,"has_summary":true,"content_length":1421,"summary_length":300,"usable_text_length":1421,"source_field":"content"},"legacy_collapsed":false,"signals":{"extract_state":"ok","extract_error":null,"extract_retries":0,"content_length":1421,"summary_length":300}},"tags":[],"format_contract_version":"news_item_formats.v1"}}}