Getting ready for the "AI & Cancer" day of Cancéropole IdF to discuss what machine learning can and cannot do in genomics.
My message: there's a lot that predictive models can do (with the usual caveats of proper evaluation), but as for interpretability, which we need for the generation of novel scientific hypotheses, robust tools with statistical guarantees are few and far between.
Will share slides later, if anyone is interested.
https://www.canceropole-idf.fr/seminaires/sem2026-ia/
#machineLearning #genomics
Chloé Azencott
npub1gf…qv9un
2026-05-21 07:08:50 UTC
Author Public Key
npub1gfxydk3rlkxae42gdm3e80zln7d428t27987lv8e9rcyjzx82acseqv9unSeen on
wss://relay.ditto.pubPublished at
2026-05-21 07:08:50 UTCEvent JSON
{
"id": "d4927e0af3c9e0e7f9cfdebb6ab8f2d1322f0330fffa4bfbf2387c57c44f924f",
"pubkey": "424c46da23fd8ddcd5486ee393bc5f9f9b551d6af14fefb0f928f04908c75771",
"created_at": 1779347330,
"kind": 1,
"tags": [
[
"t",
"MachineLearning"
],
[
"t",
"genomics"
],
[
"proxy",
"https://lipn.info/users/cazencott/statuses/116611306649842377",
"activitypub"
],
[
"client",
"Mostr",
"31990:6be38f8c63df7dbf84db7ec4a6e6fbbd8d19dca3b980efad18585c46f04b26f9:mostr",
"wss://relay.ditto.pub"
]
],
"content": "Getting ready for the \"AI \u0026 Cancer\" day of Cancéropole IdF to discuss what machine learning can and cannot do in genomics.\n\nMy message: there's a lot that predictive models can do (with the usual caveats of proper evaluation), but as for interpretability, which we need for the generation of novel scientific hypotheses, robust tools with statistical guarantees are few and far between.\n\nWill share slides later, if anyone is interested.\n \nhttps://www.canceropole-idf.fr/seminaires/sem2026-ia/\n\n#machineLearning #genomics",
"sig": "0eefa0c01ab0aa7c1b095de3e13b4dc5730e8c72938d0f147bb67e52187051a27a478abaaa3c0d5ae1c4e2569c365671a8021745f524f0555702e173b9f372a7"
}