Experimental browser for the Atmosphere
We've been interested in using Baker Lab's ProteinMPNN to design/diversify mAb seqs w/ structure data. Unfortunately its output seqs aren't very antibody-like & don't refold into the target conformation as often as we'd like
May 15, 2025, 4:30 PM
{
"text": "We've been interested in using Baker Lab's ProteinMPNN to design/diversify mAb seqs w/ structure data. Unfortunately its output seqs aren't very antibody-like & don't refold into the target conformation as often as we'd like",
"$type": "app.bsky.feed.post",
"embed": {
"$type": "app.bsky.embed.images",
"images": [
{
"alt": "Figure 1. ProteinMPNN does not generate antibody-like sequences unless ensembled with the\nantibody language model AbLang. A) V-gene sequence identity of all designs to the most similar\ngermline. Therapeutic antibody sequences were fetched from TheraSAbDaB on X February 2025. B)\nRecovery of the starting sequence. C) Recovery of antigen-binding residues in the CDRs. Y -axis\ndenotes frequency. Top to bottom: ProteinMPNN, AbMPNN, ProteinMPNN+AbLang, AbLang, ESM.\nD) Recovery of the starting conformation of CDRH3. Mann-Whitney U test results: n/s: not significant;\n*: 0.05≤p<0.5; **: 0.005≤p<0.05; ***: p<0.005.",
"image": {
"$type": "blob",
"ref": {
"$link": "bafkreidu55etii2whdvzp5e3l52xulwgtlkrf45xs2d3674gtn2tazb62e"
},
"mimeType": "image/jpeg",
"size": 465191
},
"aspectRatio": {
"width": 1892,
"height": 870
}
}
]
},
"langs": [
"en"
],
"reply": {
"root": {
"cid": "bafyreice3gjxwhqqvosjj3ljzjgsuounsbynhhoziiuca4arabekevwute",
"uri": "at://did:plc:it7vw4dejlm32tzqsdqzagd4/app.bsky.feed.post/3lp7wr4buxc2f"
},
"parent": {
"cid": "bafyreice3gjxwhqqvosjj3ljzjgsuounsbynhhoziiuca4arabekevwute",
"uri": "at://did:plc:it7vw4dejlm32tzqsdqzagd4/app.bsky.feed.post/3lp7wr4buxc2f"
}
},
"createdAt": "2025-05-15T16:30:14.106Z"
}