From owner-chemistry@ccl.net Wed Jun 26 12:14:00 2024 From: "Oleg Trott trott : caa.columbia.edu" To: CCL Subject: CCL: AlphaFold 3 vs AutoDock Vina Message-Id: <-55176-240626105239-12884-qbNT2g9hLrujTYVq+9/ajg||server.ccl.net> X-Original-From: Oleg Trott Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset="UTF-8" Date: Wed, 26 Jun 2024 07:52:12 -0700 MIME-Version: 1.0 Sent to CCL by: Oleg Trott [trott:+:caa.columbia.edu] Thanks! I hope someone re-runs DeepMind's calculations, with the same settings, but using k-fold cross-validation (like what I did at Columbia), where test and training sets are never related. This will show how well (or poorly, as the case may be) the method generalizes to unrelated proteins. ===== Oleg Trott, PhD https://olegtrott.com On Tue, Jun 25, 2024 at 11:31 AM Aydin Manzouri aydin.manzouri ~~ gmail.com wrote: > > Truly insightful. Thanks a lot. > > On Tue, Jun 25, 2024 at 1:59 AM Oleg Trott trott[-]caa.columbia.edu wrote: >> >> >> Sent to CCL by: "Oleg Trott" [trott-$-caa.columbia.edu] >> Hello, everyone! >> >> DeepMind's new AlphaFold 3 attempts to predict protein-ligand binding, and >> their publication compares it to AutoDock Vina (which I built). But their >> methodology seems strange. >> >> I wrote up my comments in a blog post. If you have an interest in AI and/or >> docking, I hope you'll find it insightful. >> >> https://olegtrott.substack.com/p/are-alphafolds-new-results-a-miracle>> E-mail to subscribers: CHEMISTRY*_*ccl.net or use:>> >> E-mail to administrators: CHEMISTRY-REQUEST*_*ccl.net or use>> >> From owner-chemistry@ccl.net Wed Jun 26 14:34:00 2024 From: "Bennion, Brian bennion1]|[llnl.gov" To: CCL Subject: CCL: AlphaFold 3 vs AutoDock Vina Message-Id: <-55177-240626143241-14315-m8K/YQXcsG1TlLeMyntT7A/a\server.ccl.net> X-Original-From: "Bennion, Brian" Content-Language: en-US Content-Type: multipart/alternative; boundary="_000_BY3PR09MB81310C98B32822DDF5BAB344F6D62BY3PR09MB8131namp_" Date: Wed, 26 Jun 2024 18:32:21 +0000 MIME-Version: 1.0 Sent to CCL by: "Bennion, Brian" [bennion1]![llnl.gov] --_000_BY3PR09MB81310C98B32822DDF5BAB344F6D62BY3PR09MB8131namp_ Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: base64 SSBhbSBmYWlybHkgbmV3IGluIHRoaXMgYXJlYSBzbyBteSBhcG9sb2dpZXMuDQpXaWxsIGstZm9s ZCBjcm9zcyB2YWxpZGF0aW9uIHNwbGl0IGRhdGFzZXRzIGJ5IHNpbWlsYXJpdHkgYXMgd2VsbCBh cyBpZGVudGl0eT8gIFRoZSBleGFtcGxlIGluIHlvdXIgcG9zdCB3aGlsZSB2ZXJ5IGhlbHBmdWwg dG8gdW5kZXJzdGFuZCB5b3VyIHBvaW50IGlzIGhpZ2hseSBzaW1wbGlmaWVkLg0KDQpXb3VsZCBD IGFuZCBDJyBmb3IgaW5zdGFuY2UgYmUgc2VwZXJhdGVkIGludG8gZGlmZmVyZW50IHNldHMgYXMg d2VsbCBhcyBtdWx0aXBsZSBpbnN0YW5jZXMgb2YgQz8NCg0KU2NhZmZvbGQgc3BsaXR0aW5nIHdo ZW4gZG9uZSBwcm9wZXJseSB3aWxsIGVuc3VyZSB0aGF0IGhpZ2hseSBzaW1pbGFyICh0YW5pbW90 byBzY29yZXMpIGNvbXBvdW5kcyBhcmUgbm90IGZvdW5kIGluIHRyYWluaW5nL3ZhbGlkYXRpb24v dGVzdCBzZXRzLg0KDQpUaGFua3MgZm9yIGFueSBoZWxwIHVuZGVyc3RhbmRpbmcgdGhpcyBhc3Bl Y3Qgb2YgdGhlIGNvbnZlcnNhdGlvbi4NCg0KYnJpYW4NCg0KDQpfX19fX19fX19fX19fX19fX19f X19fX19fX19fX19fXw0KRnJvbTogb3duZXItY2hlbWlzdHJ5K2Jlbm5pb24xPT1sbG5sLmdvdkBj Y2wubmV0IDxvd25lci1jaGVtaXN0cnkrYmVubmlvbjE9PWxsbmwuZ292QGNjbC5uZXQ+IG9uIGJl aGFsZiBvZiBPbGVnIFRyb3R0IHRyb3R0IDogY2FhLmNvbHVtYmlhLmVkdSA8b3duZXItY2hlbWlz dHJ5QGNjbC5uZXQ+DQpTZW50OiBXZWRuZXNkYXksIEp1bmUgMjYsIDIwMjQgNzo1MiBBTQ0KVG86 IEJlbm5pb24sIEJyaWFuIDxiZW5uaW9uMUBsbG5sLmdvdj4NClN1YmplY3Q6IENDTDogQWxwaGFG b2xkIDMgdnMgQXV0b0RvY2sgVmluYQ0KDQoNClNlbnQgdG8gQ0NMIGJ5OiBPbGVnIFRyb3R0IFt0 cm90dDorOmNhYS5jb2x1bWJpYS5lZHVdDQpUaGFua3MhDQoNCkkgaG9wZSBzb21lb25lIHJlLXJ1 bnMgRGVlcE1pbmQncyBjYWxjdWxhdGlvbnMsIHdpdGggdGhlIHNhbWUNCnNldHRpbmdzLCBidXQg dXNpbmcgay1mb2xkIGNyb3NzLXZhbGlkYXRpb24gKGxpa2Ugd2hhdCBJIGRpZCBhdA0KQ29sdW1i aWEpLCB3aGVyZSB0ZXN0IGFuZCB0cmFpbmluZyBzZXRzIGFyZSBuZXZlciByZWxhdGVkLiBUaGlz IHdpbGwNCnNob3cgaG93IHdlbGwgKG9yIHBvb3JseSwgYXMgdGhlIGNhc2UgbWF5IGJlKSB0aGUg bWV0aG9kIGdlbmVyYWxpemVzDQp0byB1bnJlbGF0ZWQgcHJvdGVpbnMuDQoNCg0KPT09PT0NCk9s ZWcgVHJvdHQsIFBoRA0KaHR0cHM6Ly91cmxkZWZlbnNlLnVzL3YzL19faHR0cHM6Ly9vbGVndHJv dHQuY29tX187ISFHMmtwTTd1TS1UeklGY2h1ITNwOVRkU2tQdmxXWTdzcExDemRydzhBU0hiVXZq alFrWnBTUEpnWFFvdDV1cmFTdWFLV3VNc2lsYS1aakhLQkM3LW05MmZKUXB6bUI2WVJKM0xBWnQ0 R1pHQSQNCg0KDQoNCg0KDQpPbiBUdWUsIEp1biAyNSwgMjAyNCBhdCAxMTozMeKAr0FNIEF5ZGlu IE1hbnpvdXJpIGF5ZGluLm1hbnpvdXJpIH5+DQpnbWFpbC5jb20gPG93bmVyLWNoZW1pc3RyeV8t X2NjbC5uZXQ+IHdyb3RlOg0KPg0KPiBUcnVseSBpbnNpZ2h0ZnVsLiBUaGFua3MgYSBsb3QuDQo+ DQo+IE9uIFR1ZSwgSnVuIDI1LCAyMDI0IGF0IDE6NTnigK9BTSBPbGVnIFRyb3R0IHRyb3R0Wy1d Y2FhLmNvbHVtYmlhLmVkdSA8b3duZXItY2hlbWlzdHJ5Kl8qY2NsLm5ldD4gd3JvdGU6DQo+Pg0K Pj4NCj4+IFNlbnQgdG8gQ0NMIGJ5OiAiT2xlZyAgVHJvdHQiIFt0cm90dC0kLWNhYS5jb2x1bWJp YS5lZHVdDQo+PiBIZWxsbywgZXZlcnlvbmUhDQo+Pg0KPj4gRGVlcE1pbmQncyBuZXcgQWxwaGFG b2xkIDMgYXR0ZW1wdHMgdG8gcHJlZGljdCBwcm90ZWluLWxpZ2FuZCBiaW5kaW5nLCBhbmQNCj4+ IHRoZWlyIHB1YmxpY2F0aW9uIGNvbXBhcmVzIGl0IHRvIEF1dG9Eb2NrIFZpbmEgKHdoaWNoIEkg YnVpbHQpLiBCdXQgdGhlaXINCj4+IG1ldGhvZG9sb2d5IHNlZW1zIHN0cmFuZ2UuDQo+Pg0KPj4g SSB3cm90ZSB1cCBteSBjb21tZW50cyBpbiBhIGJsb2cgcG9zdC4gSWYgeW91IGhhdmUgYW4gaW50 ZXJlc3QgaW4gQUkgYW5kL29yDQo+PiBkb2NraW5nLCBJIGhvcGUgeW91J2xsIGZpbmQgaXQgaW5z aWdodGZ1bC4NCj4+DQo+PiBodHRwczovL3VybGRlZmVuc2UudXMvdjMvX19odHRwczovL29sZWd0 cm90dC5zdWJzdGFjay5jb20vcC9hcmUtYWxwaGFmb2xkcy1uZXctcmVzdWx0cy1hLW1pcmFjbGVf XzshIUcya3BNN3VNLVR6SUZjaHUhM3A5VGRTa1B2bFdZN3NwTEN6ZHJ3OEFTSGJVdmpqUWtacFNQ SmdYUW90NXVyYVN1YUtXdU1zaWxhLVpqSEtCQzctbTkyZkpRcHptQjZZUkozTEFoaldVVzJRJCA+ PiBFLW1haWwgdG8gc3Vic2NyaWJlcnM6IENIRU1JU1RSWSpfKmNjbC5uZXQgb3IgdXNlOj4+DQo+ PiBFLW1haWwgdG8gYWRtaW5pc3RyYXRvcnM6IENIRU1JU1RSWS1SRVFVRVNUKl8qY2NsLm5ldCBv ciB1c2U+Pg0KPj4NCg0KDQoNCi09IFRoaXMgaXMgYXV0b21hdGljYWxseSBhZGRlZCB0byBlYWNo IG1lc3NhZ2UgYnkgdGhlIG1haWxpbmcgc2NyaXB0ID0tDQpUbyByZWNvdmVyIHRoZSBlbWFpbCBh ZGRyZXNzIG9mIHRoZSBhdXRob3Igb2YgdGhlIG1lc3NhZ2UsIHBsZWFzZSBjaGFuZ2UNCnRoZSBz dHJhbmdlIGNoYXJhY3RlcnMgb24gdGhlIHRvcCBsaW5lIHRvIHRoZSBAIHNpZ24uIFlvdSBjYW4g YWxzbw0KbG9vayB1cCB0aGUgWC1PcmlnaW5hbC1Gcm9tOiBsaW5lIGluIHRoZSBtYWlsIGhlYWRl ci4NCg0KRS1tYWlsIHRvIHN1YnNjcmliZXJzOiBDSEVNSVNUUllAY2NsLm5ldCBvciB1c2U6DQog ICAgICBodHRwczovL3VybGRlZmVuc2UudXMvdjMvX19odHRwOi8vd3d3LmNjbC5uZXQvY2dpLWJp bi9jY2wvc2VuZF9jY2xfbWVzc2FnZV9fOyEhRzJrcE03dU0tVHpJRmNodSEzcDlUZFNrUHZsV1k3 c3BMQ3pkcnc4QVNIYlV2ampRa1pwU1BKZ1hRb3Q1dXJhU3VhS1d1TXNpbGEtWmpIS0JDNy1tOTJm SlFwem1CNllSSjNMREtRQVI2VUEkDQoNCkUtbWFpbCB0byBhZG1pbmlzdHJhdG9yczogQ0hFTUlT VFJZLVJFUVVFU1RAY2NsLm5ldCBvciB1c2UNCiAgICAgIGh0dHBzOi8vdXJsZGVmZW5zZS51cy92 My9fX2h0dHA6Ly93d3cuY2NsLm5ldC9jZ2ktYmluL2NjbC9zZW5kX2NjbF9tZXNzYWdlX187ISFH MmtwTTd1TS1UeklGY2h1ITNwOVRkU2tQdmxXWTdzcExDemRydzhBU0hiVXZqalFrWnBTUEpnWFFv dDV1cmFTdWFLV3VNc2lsYS1aakhLQkM3LW05MmZKUXB6bUI2WVJKM0xES1FBUjZVQSQNCg0KU3Vi c2NyaWJlL1Vuc3Vic2NyaWJlOg0KICAgICAgaHR0cHM6Ly91cmxkZWZlbnNlLnVzL3YzL19faHR0 cDovL3d3dy5jY2wubmV0L2NoZW1pc3RyeS9zdWJfdW5zdWIuc2h0bWxfXzshIUcya3BNN3VNLVR6 SUZjaHUhM3A5VGRTa1B2bFdZN3NwTEN6ZHJ3OEFTSGJVdmpqUWtacFNQSmdYUW90NXVyYVN1YUtX dU1zaWxhLVpqSEtCQzctbTkyZkpRcHptQjZZUkozTENQV1cyQ1Z3JA0KDQpCZWZvcmUgcG9zdGlu ZywgY2hlY2sgd2FpdCB0aW1lIGF0OiBodHRwczovL3VybGRlZmVuc2UudXMvdjMvX19odHRwOi8v d3d3LmNjbC5uZXRfXzshIUcya3BNN3VNLVR6SUZjaHUhM3A5VGRTa1B2bFdZN3NwTEN6ZHJ3OEFT SGJVdmpqUWtacFNQSmdYUW90NXVyYVN1YUtXdU1zaWxhLVpqSEtCQzctbTkyZkpRcHptQjZZUkoz TERscFU4MzBRJA0KDQpKb2I6IGh0dHBzOi8vdXJsZGVmZW5zZS51cy92My9fX2h0dHA6Ly93d3cu Y2NsLm5ldC9qb2JzX187ISFHMmtwTTd1TS1UeklGY2h1ITNwOVRkU2tQdmxXWTdzcExDemRydzhB U0hiVXZqalFrWnBTUEpnWFFvdDV1cmFTdWFLV3VNc2lsYS1aakhLQkM3LW05MmZKUXB6bUI2WVJK M0xEQ1RLV2pBUSQNCkNvbmZlcmVuY2VzOiBodHRwczovL3VybGRlZmVuc2UudXMvdjMvX19odHRw Oi8vc2VydmVyLmNjbC5uZXQvY2hlbWlzdHJ5L2Fubm91bmNlbWVudHMvY29uZmVyZW5jZXMvX187 ISFHMmtwTTd1TS1UeklGY2h1ITNwOVRkU2tQdmxXWTdzcExDemRydzhBU0hiVXZqalFrWnBTUEpn WFFvdDV1cmFTdWFLV3VNc2lsYS1aakhLQkM3LW05MmZKUXB6bUI2WVJKM0xCWGxVcWx5ZyQNCg0K U2VhcmNoIE1lc3NhZ2VzOiBodHRwczovL3VybGRlZmVuc2UudXMvdjMvX19odHRwOi8vd3d3LmNj bC5uZXQvY2hlbWlzdHJ5L3NlYXJjaGNjbC9pbmRleC5zaHRtbF9fOyEhRzJrcE03dU0tVHpJRmNo dSEzcDlUZFNrUHZsV1k3c3BMQ3pkcnc4QVNIYlV2ampRa1pwU1BKZ1hRb3Q1dXJhU3VhS1d1TXNp bGEtWmpIS0JDNy1tOTJmSlFwem1CNllSSjNMQmYwUFFBaXckDQoNCklmIHlvdXIgbWFpbCBib3Vu Y2VzIGZyb20gQ0NMIHdpdGggNS43LjEgZXJyb3IsIGNoZWNrOg0KICAgICAgaHR0cHM6Ly91cmxk ZWZlbnNlLnVzL3YzL19faHR0cDovL3d3dy5jY2wubmV0L3NwYW1tZXJzLnR4dF9fOyEhRzJrcE03 dU0tVHpJRmNodSEzcDlUZFNrUHZsV1k3c3BMQ3pkcnc4QVNIYlV2ampRa1pwU1BKZ1hRb3Q1dXJh U3VhS1d1TXNpbGEtWmpIS0JDNy1tOTJmSlFwem1CNllSSjNMQ1BOSEJyVXckDQoNClJURkk6IGh0 dHBzOi8vdXJsZGVmZW5zZS51cy92My9fX2h0dHA6Ly93d3cuY2NsLm5ldC9jaGVtaXN0cnkvYWJv dXRjY2wvaW5zdHJ1Y3Rpb25zL19fOyEhRzJrcE03dU0tVHpJRmNodSEzcDlUZFNrUHZsV1k3c3BM Q3pkcnc4QVNIYlV2ampRa1pwU1BKZ1hRb3Q1dXJhU3VhS1d1TXNpbGEtWmpIS0JDNy1tOTJmSlFw em1CNllSSjNMRFpIck11dFEkDQoNCg0K --_000_BY3PR09MB81310C98B32822DDF5BAB344F6D62BY3PR09MB8131namp_ Content-Type: text/html; charset="utf-8" Content-Transfer-Encoding: base64 PGh0bWw+DQo8aGVhZD4NCjxtZXRhIGh0dHAtZXF1aXY9IkNvbnRlbnQtVHlwZSIgY29udGVudD0i dGV4dC9odG1sOyBjaGFyc2V0PXV0Zi04Ij4NCjxzdHlsZSB0eXBlPSJ0ZXh0L2NzcyIgc3R5bGU9 ImRpc3BsYXk6bm9uZTsiPiBQIHttYXJnaW4tdG9wOjA7bWFyZ2luLWJvdHRvbTowO30gPC9zdHls ZT4NCjwvaGVhZD4NCjxib2R5IGRpcj0ibHRyIj4NCjxkaXYgY2xhc3M9ImVsZW1lbnRUb1Byb29m IiBzdHlsZT0iZm9udC1mYW1pbHk6IEFwdG9zLCBBcHRvc19FbWJlZGRlZEZvbnQsIEFwdG9zX01T Rm9udFNlcnZpY2UsIENhbGlicmksIEhlbHZldGljYSwgc2Fucy1zZXJpZjsgZm9udC1zaXplOiAx MXB0OyBjb2xvcjogcmdiKDAsIDAsIDApOyI+DQpJIGFtIGZhaXJseSBuZXcgaW4gdGhpcyBhcmVh IHNvIG15IGFwb2xvZ2llcy48L2Rpdj4NCjxkaXYgY2xhc3M9ImVsZW1lbnRUb1Byb29mIiBzdHls ZT0iZm9udC1mYW1pbHk6IEFwdG9zLCBBcHRvc19FbWJlZGRlZEZvbnQsIEFwdG9zX01TRm9udFNl cnZpY2UsIENhbGlicmksIEhlbHZldGljYSwgc2Fucy1zZXJpZjsgZm9udC1zaXplOiAxMXB0OyBj b2xvcjogcmdiKDAsIDAsIDApOyI+DQpXaWxsIGstZm9sZCBjcm9zcyB2YWxpZGF0aW9uIHNwbGl0 IGRhdGFzZXRzIGJ5IHNpbWlsYXJpdHkgYXMgd2VsbCBhcyBpZGVudGl0eT8mbmJzcDsgVGhlIGV4 YW1wbGUgaW4geW91ciBwb3N0IHdoaWxlIHZlcnkgaGVscGZ1bCB0byB1bmRlcnN0YW5kIHlvdXIg cG9pbnQgaXMgaGlnaGx5IHNpbXBsaWZpZWQuPC9kaXY+DQo8ZGl2IGNsYXNzPSJlbGVtZW50VG9Q cm9vZiIgc3R5bGU9ImZvbnQtZmFtaWx5OiBBcHRvcywgQXB0b3NfRW1iZWRkZWRGb250LCBBcHRv c19NU0ZvbnRTZXJ2aWNlLCBDYWxpYnJpLCBIZWx2ZXRpY2EsIHNhbnMtc2VyaWY7IGZvbnQtc2l6 ZTogMTFwdDsgY29sb3I6IHJnYigwLCAwLCAwKTsiPg0KPGJyPg0KPC9kaXY+DQo8ZGl2IGNsYXNz PSJlbGVtZW50VG9Qcm9vZiIgc3R5bGU9ImZvbnQtZmFtaWx5OiBBcHRvcywgQXB0b3NfRW1iZWRk ZWRGb250LCBBcHRvc19NU0ZvbnRTZXJ2aWNlLCBDYWxpYnJpLCBIZWx2ZXRpY2EsIHNhbnMtc2Vy aWY7IGZvbnQtc2l6ZTogMTFwdDsgY29sb3I6IHJnYigwLCAwLCAwKTsiPg0KV291bGQgQyBhbmQg QycgZm9yIGluc3RhbmNlIGJlIHNlcGVyYXRlZCBpbnRvIGRpZmZlcmVudCBzZXRzIGFzIHdlbGwg YXMgbXVsdGlwbGUgaW5zdGFuY2VzIG9mIEM/PC9kaXY+DQo8ZGl2IGNsYXNzPSJlbGVtZW50VG9Q cm9vZiIgc3R5bGU9ImZvbnQtZmFtaWx5OiBBcHRvcywgQXB0b3NfRW1iZWRkZWRGb250LCBBcHRv c19NU0ZvbnRTZXJ2aWNlLCBDYWxpYnJpLCBIZWx2ZXRpY2EsIHNhbnMtc2VyaWY7IGZvbnQtc2l6 ZTogMTFwdDsgY29sb3I6IHJnYigwLCAwLCAwKTsiPg0KPGJyPg0KPC9kaXY+DQo8ZGl2IGNsYXNz PSJlbGVtZW50VG9Qcm9vZiIgc3R5bGU9ImZvbnQtZmFtaWx5OiBBcHRvcywgQXB0b3NfRW1iZWRk ZWRGb250LCBBcHRvc19NU0ZvbnRTZXJ2aWNlLCBDYWxpYnJpLCBIZWx2ZXRpY2EsIHNhbnMtc2Vy aWY7IGZvbnQtc2l6ZTogMTFwdDsgY29sb3I6IHJnYigwLCAwLCAwKTsiPg0KU2NhZmZvbGQgc3Bs aXR0aW5nIHdoZW4gZG9uZSBwcm9wZXJseSB3aWxsIGVuc3VyZSB0aGF0IGhpZ2hseSBzaW1pbGFy ICh0YW5pbW90byBzY29yZXMpIGNvbXBvdW5kcyBhcmUgbm90IGZvdW5kIGluIHRyYWluaW5nL3Zh bGlkYXRpb24vdGVzdCBzZXRzLjwvZGl2Pg0KPGRpdiBjbGFzcz0iZWxlbWVudFRvUHJvb2YiIHN0 eWxlPSJmb250LWZhbWlseTogQXB0b3MsIEFwdG9zX0VtYmVkZGVkRm9udCwgQXB0b3NfTVNGb250 U2VydmljZSwgQ2FsaWJyaSwgSGVsdmV0aWNhLCBzYW5zLXNlcmlmOyBmb250LXNpemU6IDExcHQ7 IGNvbG9yOiByZ2IoMCwgMCwgMCk7Ij4NCjxicj4NCjwvZGl2Pg0KPGRpdiBjbGFzcz0iZWxlbWVu dFRvUHJvb2YiIHN0eWxlPSJmb250LWZhbWlseTogQXB0b3MsIEFwdG9zX0VtYmVkZGVkRm9udCwg QXB0b3NfTVNGb250U2VydmljZSwgQ2FsaWJyaSwgSGVsdmV0aWNhLCBzYW5zLXNlcmlmOyBmb250 LXNpemU6IDExcHQ7IGNvbG9yOiByZ2IoMCwgMCwgMCk7Ij4NClRoYW5rcyBmb3IgYW55IGhlbHAg dW5kZXJzdGFuZGluZyB0aGlzIGFzcGVjdCBvZiB0aGUgY29udmVyc2F0aW9uLjwvZGl2Pg0KPGRp diBjbGFzcz0iZWxlbWVudFRvUHJvb2YiIHN0eWxlPSJmb250LWZhbWlseTogQXB0b3MsIEFwdG9z X0VtYmVkZGVkRm9udCwgQXB0b3NfTVNGb250U2VydmljZSwgQ2FsaWJyaSwgSGVsdmV0aWNhLCBz YW5zLXNlcmlmOyBmb250LXNpemU6IDExcHQ7IGNvbG9yOiByZ2IoMCwgMCwgMCk7Ij4NCjxicj4N CjwvZGl2Pg0KPGRpdiBjbGFzcz0iZWxlbWVudFRvUHJvb2YiIHN0eWxlPSJmb250LWZhbWlseTog QXB0b3MsIEFwdG9zX0VtYmVkZGVkRm9udCwgQXB0b3NfTVNGb250U2VydmljZSwgQ2FsaWJyaSwg SGVsdmV0aWNhLCBzYW5zLXNlcmlmOyBmb250LXNpemU6IDExcHQ7IGNvbG9yOiByZ2IoMCwgMCwg MCk7Ij4NCmJyaWFuPC9kaXY+DQo8ZGl2IGNsYXNzPSJlbGVtZW50VG9Qcm9vZiIgc3R5bGU9ImZv bnQtZmFtaWx5OiBBcHRvcywgQXB0b3NfRW1iZWRkZWRGb250LCBBcHRvc19NU0ZvbnRTZXJ2aWNl LCBDYWxpYnJpLCBIZWx2ZXRpY2EsIHNhbnMtc2VyaWY7IGZvbnQtc2l6ZTogMTFwdDsgY29sb3I6 IHJnYigwLCAwLCAwKTsiPg0KPGJyPg0KPC9kaXY+DQo8ZGl2IGlkPSJhcHBlbmRvbnNlbmQiPjwv ZGl2Pg0KPGRpdiBzdHlsZT0iZm9udC1mYW1pbHk6IEFwdG9zLCBBcHRvc19FbWJlZGRlZEZvbnQs IEFwdG9zX01TRm9udFNlcnZpY2UsIENhbGlicmksIEhlbHZldGljYSwgc2Fucy1zZXJpZjsgZm9u dC1zaXplOiAxMXB0OyBjb2xvcjogcmdiKDAsIDAsIDApOyI+DQo8YnI+DQo8L2Rpdj4NCjxociBz dHlsZT0iZGlzcGxheTogaW5saW5lLWJsb2NrOyB3aWR0aDogOTglOyI+DQo8ZGl2IGlkPSJkaXZS cGx5RndkTXNnIiBkaXI9Imx0ciI+PHNwYW4gc3R5bGU9ImZvbnQtZmFtaWx5OiBDYWxpYnJpLCBz YW5zLXNlcmlmOyBmb250LXNpemU6IDExcHQ7IGNvbG9yOiByZ2IoMCwgMCwgMCk7Ij48Yj5Gcm9t OjwvYj4mbmJzcDtvd25lci1jaGVtaXN0cnkrYmVubmlvbjE9PWxsbmwuZ292QGNjbC5uZXQgJmx0 O293bmVyLWNoZW1pc3RyeStiZW5uaW9uMT09bGxubC5nb3ZAY2NsLm5ldCZndDsgb24gYmVoYWxm IG9mIE9sZWcgVHJvdHQgdHJvdHQgOiBjYWEuY29sdW1iaWEuZWR1DQogJmx0O293bmVyLWNoZW1p c3RyeUBjY2wubmV0Jmd0Ozxicj4NCjxiPlNlbnQ6PC9iPiZuYnNwO1dlZG5lc2RheSwgSnVuZSAy NiwgMjAyNCA3OjUyIEFNPGJyPg0KPGI+VG86PC9iPiZuYnNwO0Jlbm5pb24sIEJyaWFuICZsdDti ZW5uaW9uMUBsbG5sLmdvdiZndDs8YnI+DQo8Yj5TdWJqZWN0OjwvYj4mbmJzcDtDQ0w6IEFscGhh Rm9sZCAzIHZzIEF1dG9Eb2NrIFZpbmE8L3NwYW4+DQo8ZGl2PiZuYnNwOzwvZGl2Pg0KPC9kaXY+ DQo8ZGl2IHN0eWxlPSJmb250LXNpemU6IDExcHQ7Ij48YnI+DQpTZW50IHRvIENDTCBieTogT2xl ZyBUcm90dCBbdHJvdHQ6KzpjYWEuY29sdW1iaWEuZWR1XTxicj4NClRoYW5rcyE8YnI+DQo8YnI+ DQpJIGhvcGUgc29tZW9uZSByZS1ydW5zIERlZXBNaW5kJ3MgY2FsY3VsYXRpb25zLCB3aXRoIHRo ZSBzYW1lPGJyPg0Kc2V0dGluZ3MsIGJ1dCB1c2luZyBrLWZvbGQgY3Jvc3MtdmFsaWRhdGlvbiAo bGlrZSB3aGF0IEkgZGlkIGF0PGJyPg0KQ29sdW1iaWEpLCB3aGVyZSB0ZXN0IGFuZCB0cmFpbmlu ZyBzZXRzIGFyZSBuZXZlciByZWxhdGVkLiBUaGlzIHdpbGw8YnI+DQpzaG93IGhvdyB3ZWxsIChv ciBwb29ybHksIGFzIHRoZSBjYXNlIG1heSBiZSkgdGhlIG1ldGhvZCBnZW5lcmFsaXplczxicj4N CnRvIHVucmVsYXRlZCBwcm90ZWlucy48YnI+DQo8YnI+DQo8YnI+DQo9PT09PTxicj4NCk9sZWcg VHJvdHQsIFBoRDxicj4NCjxhIGhyZWY9Imh0dHBzOi8vdXJsZGVmZW5zZS51cy92My9fX2h0dHBz Oi8vb2xlZ3Ryb3R0LmNvbV9fOyEhRzJrcE03dU0tVHpJRmNodSEzcDlUZFNrUHZsV1k3c3BMQ3pk cnc4QVNIYlV2ampRa1pwU1BKZ1hRb3Q1dXJhU3VhS1d1TXNpbGEtWmpIS0JDNy1tOTJmSlFwem1C NllSSjNMQVp0NEdaR0EkIiBpZD0iT1dBOTQxZDE2OGEtMDIzYS1iMjdjLWIwZjctZWI4NmQ3YjY2 MjRiIiBjbGFzcz0iT1dBQXV0b0xpbmsiIGRhdGEtYXV0aD0iTm90QXBwbGljYWJsZSI+aHR0cHM6 Ly91cmxkZWZlbnNlLnVzL3YzL19faHR0cHM6Ly9vbGVndHJvdHQuY29tX187ISFHMmtwTTd1TS1U eklGY2h1ITNwOVRkU2tQdmxXWTdzcExDemRydzhBU0hiVXZqalFrWnBTUEpnWFFvdDV1cmFTdWFL V3VNc2lsYS1aakhLQkM3LW05MmZKUXB6bUI2WVJKM0xBWnQ0R1pHQSQ8L2E+PGJyPg0KPGJyPg0K PGJyPg0KPGJyPg0KPGJyPg0KPGJyPg0KT24gVHVlLCBKdW4gMjUsIDIwMjQgYXQgMTE6MzHigK9B TSBBeWRpbiBNYW56b3VyaSBheWRpbi5tYW56b3VyaSB+fjxicj4NCmdtYWlsLmNvbSAmbHQ7b3du ZXItY2hlbWlzdHJ5Xy1fY2NsLm5ldCZndDsgd3JvdGU6PGJyPg0KJmd0Ozxicj4NCiZndDsgVHJ1 bHkgaW5zaWdodGZ1bC4gVGhhbmtzIGEgbG90Ljxicj4NCiZndDs8YnI+DQomZ3Q7IE9uIFR1ZSwg SnVuIDI1LCAyMDI0IGF0IDE6NTnigK9BTSBPbGVnIFRyb3R0IHRyb3R0Wy1dY2FhLmNvbHVtYmlh LmVkdSAmbHQ7b3duZXItY2hlbWlzdHJ5Kl8qY2NsLm5ldCZndDsgd3JvdGU6PGJyPg0KJmd0OyZn dDs8YnI+DQomZ3Q7Jmd0Ozxicj4NCiZndDsmZ3Q7IFNlbnQgdG8gQ0NMIGJ5OiAmcXVvdDtPbGVn Jm5ic3A7IFRyb3R0JnF1b3Q7IFt0cm90dC0kLWNhYS5jb2x1bWJpYS5lZHVdPGJyPg0KJmd0OyZn dDsgSGVsbG8sIGV2ZXJ5b25lITxicj4NCiZndDsmZ3Q7PGJyPg0KJmd0OyZndDsgRGVlcE1pbmQn cyBuZXcgQWxwaGFGb2xkIDMgYXR0ZW1wdHMgdG8gcHJlZGljdCBwcm90ZWluLWxpZ2FuZCBiaW5k aW5nLCBhbmQ8YnI+DQomZ3Q7Jmd0OyB0aGVpciBwdWJsaWNhdGlvbiBjb21wYXJlcyBpdCB0byBB dXRvRG9jayBWaW5hICh3aGljaCBJIGJ1aWx0KS4gQnV0IHRoZWlyPGJyPg0KJmd0OyZndDsgbWV0 aG9kb2xvZ3kgc2VlbXMgc3RyYW5nZS48YnI+DQomZ3Q7Jmd0Ozxicj4NCiZndDsmZ3Q7IEkgd3Jv dGUgdXAgbXkgY29tbWVudHMgaW4gYSBibG9nIHBvc3QuIElmIHlvdSBoYXZlIGFuIGludGVyZXN0 IGluIEFJIGFuZC9vcjxicj4NCiZndDsmZ3Q7IGRvY2tpbmcsIEkgaG9wZSB5b3UnbGwgZmluZCBp dCBpbnNpZ2h0ZnVsLjxicj4NCiZndDsmZ3Q7PGJyPg0KJmd0OyZndDsgPGEgaHJlZj0iaHR0cHM6 Ly91cmxkZWZlbnNlLnVzL3YzL19faHR0cHM6Ly9vbGVndHJvdHQuc3Vic3RhY2suY29tL3AvYXJl LWFscGhhZm9sZHMtbmV3LXJlc3VsdHMtYS1taXJhY2xlX187ISFHMmtwTTd1TS1UeklGY2h1ITNw OVRkU2tQdmxXWTdzcExDemRydzhBU0hiVXZqalFrWnBTUEpnWFFvdDV1cmFTdWFLV3VNc2lsYS1a akhLQkM3LW05MmZKUXB6bUI2WVJKM0xBaGpXVVcyUSQiIGlkPSJPV0E5NWI3ZTUzZi01NjkzLTY2 YzUtZjEwMC0xNjJiNDBjZjJiYWYiIGNsYXNzPSJPV0FBdXRvTGluayIgZGF0YS1hdXRoPSJOb3RB cHBsaWNhYmxlIj4NCmh0dHBzOi8vdXJsZGVmZW5zZS51cy92My9fX2h0dHBzOi8vb2xlZ3Ryb3R0 LnN1YnN0YWNrLmNvbS9wL2FyZS1hbHBoYWZvbGRzLW5ldy1yZXN1bHRzLWEtbWlyYWNsZV9fOyEh RzJrcE03dU0tVHpJRmNodSEzcDlUZFNrUHZsV1k3c3BMQ3pkcnc4QVNIYlV2ampRa1pwU1BKZ1hR b3Q1dXJhU3VhS1d1TXNpbGEtWmpIS0JDNy1tOTJmSlFwem1CNllSSjNMQWhqV1VXMlEkPC9hPiZu YnNwOyZndDsmZ3Q7IEUtbWFpbCB0byBzdWJzY3JpYmVyczogQ0hFTUlTVFJZKl8qY2NsLm5ldA0K IG9yIHVzZTomZ3Q7Jmd0Ozxicj4NCiZndDsmZ3Q7IEUtbWFpbCB0byBhZG1pbmlzdHJhdG9yczog Q0hFTUlTVFJZLVJFUVVFU1QqXypjY2wubmV0IG9yIHVzZSZndDsmZ3Q7PGJyPg0KJmd0OyZndDs8 YnI+DQo8YnI+DQo8YnI+DQo8YnI+DQotPSBUaGlzIGlzIGF1dG9tYXRpY2FsbHkgYWRkZWQgdG8g ZWFjaCBtZXNzYWdlIGJ5IHRoZSBtYWlsaW5nIHNjcmlwdCA9LTxicj4NClRvIHJlY292ZXIgdGhl IGVtYWlsIGFkZHJlc3Mgb2YgdGhlIGF1dGhvciBvZiB0aGUgbWVzc2FnZSwgcGxlYXNlIGNoYW5n ZTxicj4NCnRoZSBzdHJhbmdlIGNoYXJhY3RlcnMgb24gdGhlIHRvcCBsaW5lIHRvIHRoZSBAIHNp Z24uIFlvdSBjYW4gYWxzbzxicj4NCmxvb2sgdXAgdGhlIFgtT3JpZ2luYWwtRnJvbTogbGluZSBp biB0aGUgbWFpbCBoZWFkZXIuPGJyPg0KPGJyPg0KRS1tYWlsIHRvIHN1YnNjcmliZXJzOiBDSEVN SVNUUllAY2NsLm5ldCBvciB1c2U6PGJyPg0KJm5ic3A7Jm5ic3A7Jm5ic3A7Jm5ic3A7Jm5ic3A7 IDxhIGhyZWY9Imh0dHBzOi8vdXJsZGVmZW5zZS51cy92My9fX2h0dHA6Ly93d3cuY2NsLm5ldC9j Z2ktYmluL2NjbC9zZW5kX2NjbF9tZXNzYWdlX187ISFHMmtwTTd1TS1UeklGY2h1ITNwOVRkU2tQ dmxXWTdzcExDemRydzhBU0hiVXZqalFrWnBTUEpnWFFvdDV1cmFTdWFLV3VNc2lsYS1aakhLQkM3 LW05MmZKUXB6bUI2WVJKM0xES1FBUjZVQSQiIGlkPSJPV0FlZTljMjQzZi1mYmUwLWM3YzEtMGE3 NS0yY2RhMWRlMzdiYTMiIGNsYXNzPSJPV0FBdXRvTGluayIgZGF0YS1hdXRoPSJOb3RBcHBsaWNh YmxlIj4NCmh0dHBzOi8vdXJsZGVmZW5zZS51cy92My9fX2h0dHA6Ly93d3cuY2NsLm5ldC9jZ2kt YmluL2NjbC9zZW5kX2NjbF9tZXNzYWdlX187ISFHMmtwTTd1TS1UeklGY2h1ITNwOVRkU2tQdmxX WTdzcExDemRydzhBU0hiVXZqalFrWnBTUEpnWFFvdDV1cmFTdWFLV3VNc2lsYS1aakhLQkM3LW05 MmZKUXB6bUI2WVJKM0xES1FBUjZVQSQ8L2E+PGJyPg0KPGJyPg0KRS1tYWlsIHRvIGFkbWluaXN0 cmF0b3JzOiBDSEVNSVNUUlktUkVRVUVTVEBjY2wubmV0IG9yIHVzZTxicj4NCiZuYnNwOyZuYnNw OyZuYnNwOyZuYnNwOyZuYnNwOyA8YSBocmVmPSJodHRwczovL3VybGRlZmVuc2UudXMvdjMvX19o dHRwOi8vd3d3LmNjbC5uZXQvY2dpLWJpbi9jY2wvc2VuZF9jY2xfbWVzc2FnZV9fOyEhRzJrcE03 dU0tVHpJRmNodSEzcDlUZFNrUHZsV1k3c3BMQ3pkcnc4QVNIYlV2ampRa1pwU1BKZ1hRb3Q1dXJh U3VhS1d1TXNpbGEtWmpIS0JDNy1tOTJmSlFwem1CNllSSjNMREtRQVI2VUEkIiBpZD0iT1dBNzM3 NDBhNGEtNjAxNS1kZDJjLTI0NGItYmU0NjExM2U1ZDhhIiBjbGFzcz0iT1dBQXV0b0xpbmsiIGRh dGEtYXV0aD0iTm90QXBwbGljYWJsZSI+DQpodHRwczovL3VybGRlZmVuc2UudXMvdjMvX19odHRw Oi8vd3d3LmNjbC5uZXQvY2dpLWJpbi9jY2wvc2VuZF9jY2xfbWVzc2FnZV9fOyEhRzJrcE03dU0t VHpJRmNodSEzcDlUZFNrUHZsV1k3c3BMQ3pkcnc4QVNIYlV2ampRa1pwU1BKZ1hRb3Q1dXJhU3Vh S1d1TXNpbGEtWmpIS0JDNy1tOTJmSlFwem1CNllSSjNMREtRQVI2VUEkPC9hPjxicj4NCjxicj4N ClN1YnNjcmliZS9VbnN1YnNjcmliZTo8YnI+DQombmJzcDsmbmJzcDsmbmJzcDsmbmJzcDsmbmJz cDsgPGEgaHJlZj0iaHR0cHM6Ly91cmxkZWZlbnNlLnVzL3YzL19faHR0cDovL3d3dy5jY2wubmV0 L2NoZW1pc3RyeS9zdWJfdW5zdWIuc2h0bWxfXzshIUcya3BNN3VNLVR6SUZjaHUhM3A5VGRTa1B2 bFdZN3NwTEN6ZHJ3OEFTSGJVdmpqUWtacFNQSmdYUW90NXVyYVN1YUtXdU1zaWxhLVpqSEtCQzct bTkyZkpRcHptQjZZUkozTENQV1cyQ1Z3JCIgaWQ9Ik9XQTFmMjc3ZjBlLTFiOWQtOGYwYS05ZGEw LWNhN2U1ZDgzYmJmZSIgY2xhc3M9Ik9XQUF1dG9MaW5rIiBkYXRhLWF1dGg9Ik5vdEFwcGxpY2Fi bGUiPg0KaHR0cHM6Ly91cmxkZWZlbnNlLnVzL3YzL19faHR0cDovL3d3dy5jY2wubmV0L2NoZW1p c3RyeS9zdWJfdW5zdWIuc2h0bWxfXzshIUcya3BNN3VNLVR6SUZjaHUhM3A5VGRTa1B2bFdZN3Nw TEN6ZHJ3OEFTSGJVdmpqUWtacFNQSmdYUW90NXVyYVN1YUtXdU1zaWxhLVpqSEtCQzctbTkyZkpR cHptQjZZUkozTENQV1cyQ1Z3JDwvYT48YnI+DQo8YnI+DQpCZWZvcmUgcG9zdGluZywgY2hlY2sg d2FpdCB0aW1lIGF0OiA8YSBocmVmPSJodHRwczovL3VybGRlZmVuc2UudXMvdjMvX19odHRwOi8v d3d3LmNjbC5uZXRfXzshIUcya3BNN3VNLVR6SUZjaHUhM3A5VGRTa1B2bFdZN3NwTEN6ZHJ3OEFT SGJVdmpqUWtacFNQSmdYUW90NXVyYVN1YUtXdU1zaWxhLVpqSEtCQzctbTkyZkpRcHptQjZZUkoz TERscFU4MzBRJCIgaWQ9Ik9XQTg3OGJlNTMwLTM2N2EtMTNjYS0yNzZmLThmYzQ5ZGIzMTNjYyIg Y2xhc3M9Ik9XQUF1dG9MaW5rIiBkYXRhLWF1dGg9Ik5vdEFwcGxpY2FibGUiPg0KaHR0cHM6Ly91 cmxkZWZlbnNlLnVzL3YzL19faHR0cDovL3d3dy5jY2wubmV0X187ISFHMmtwTTd1TS1UeklGY2h1 ITNwOVRkU2tQdmxXWTdzcExDemRydzhBU0hiVXZqalFrWnBTUEpnWFFvdDV1cmFTdWFLV3VNc2ls YS1aakhLQkM3LW05MmZKUXB6bUI2WVJKM0xEbHBVODMwUSQ8L2E+PGJyPg0KPGJyPg0KSm9iOiA8 YSBocmVmPSJodHRwczovL3VybGRlZmVuc2UudXMvdjMvX19odHRwOi8vd3d3LmNjbC5uZXQvam9i c19fOyEhRzJrcE03dU0tVHpJRmNodSEzcDlUZFNrUHZsV1k3c3BMQ3pkcnc4QVNIYlV2ampRa1pw U1BKZ1hRb3Q1dXJhU3VhS1d1TXNpbGEtWmpIS0JDNy1tOTJmSlFwem1CNllSSjNMRENUS1dqQVEk IiBpZD0iT1dBOTAxZDJjMmQtNTVlMC0yMjY0LTc5OTAtYThkOGYwNDRkNTg4IiBjbGFzcz0iT1dB QXV0b0xpbmsiIGRhdGEtYXV0aD0iTm90QXBwbGljYWJsZSI+DQpodHRwczovL3VybGRlZmVuc2Uu dXMvdjMvX19odHRwOi8vd3d3LmNjbC5uZXQvam9ic19fOyEhRzJrcE03dU0tVHpJRmNodSEzcDlU ZFNrUHZsV1k3c3BMQ3pkcnc4QVNIYlV2ampRa1pwU1BKZ1hRb3Q1dXJhU3VhS1d1TXNpbGEtWmpI S0JDNy1tOTJmSlFwem1CNllSSjNMRENUS1dqQVEkPC9hPiZuYnNwOzxicj4NCkNvbmZlcmVuY2Vz OiA8YSBocmVmPSJodHRwczovL3VybGRlZmVuc2UudXMvdjMvX19odHRwOi8vc2VydmVyLmNjbC5u ZXQvY2hlbWlzdHJ5L2Fubm91bmNlbWVudHMvY29uZmVyZW5jZXMvX187ISFHMmtwTTd1TS1UeklG Y2h1ITNwOVRkU2tQdmxXWTdzcExDemRydzhBU0hiVXZqalFrWnBTUEpnWFFvdDV1cmFTdWFLV3VN c2lsYS1aakhLQkM3LW05MmZKUXB6bUI2WVJKM0xCWGxVcWx5ZyQiIGlkPSJPV0FlZjRiNDRiZi00 N2ZmLTIzNWEtY2NlOS0yYmVhODJiYzVkOWQiIGNsYXNzPSJPV0FBdXRvTGluayIgZGF0YS1hdXRo PSJOb3RBcHBsaWNhYmxlIj4NCmh0dHBzOi8vdXJsZGVmZW5zZS51cy92My9fX2h0dHA6Ly9zZXJ2 ZXIuY2NsLm5ldC9jaGVtaXN0cnkvYW5ub3VuY2VtZW50cy9jb25mZXJlbmNlcy9fXzshIUcya3BN N3VNLVR6SUZjaHUhM3A5VGRTa1B2bFdZN3NwTEN6ZHJ3OEFTSGJVdmpqUWtacFNQSmdYUW90NXVy YVN1YUtXdU1zaWxhLVpqSEtCQzctbTkyZkpRcHptQjZZUkozTEJYbFVxbHlnJDwvYT48YnI+DQo8 YnI+DQpTZWFyY2ggTWVzc2FnZXM6IDxhIGhyZWY9Imh0dHBzOi8vdXJsZGVmZW5zZS51cy92My9f X2h0dHA6Ly93d3cuY2NsLm5ldC9jaGVtaXN0cnkvc2VhcmNoY2NsL2luZGV4LnNodG1sX187ISFH MmtwTTd1TS1UeklGY2h1ITNwOVRkU2tQdmxXWTdzcExDemRydzhBU0hiVXZqalFrWnBTUEpnWFFv dDV1cmFTdWFLV3VNc2lsYS1aakhLQkM3LW05MmZKUXB6bUI2WVJKM0xCZjBQUUFpdyQiIGlkPSJP V0EyZTY5NzM3ZC1mZTEyLTMxNzMtZWI3MC0yYzA4MjdiMzI1MGIiIGNsYXNzPSJPV0FBdXRvTGlu ayIgZGF0YS1hdXRoPSJOb3RBcHBsaWNhYmxlIj4NCmh0dHBzOi8vdXJsZGVmZW5zZS51cy92My9f X2h0dHA6Ly93d3cuY2NsLm5ldC9jaGVtaXN0cnkvc2VhcmNoY2NsL2luZGV4LnNodG1sX187ISFH MmtwTTd1TS1UeklGY2h1ITNwOVRkU2tQdmxXWTdzcExDemRydzhBU0hiVXZqalFrWnBTUEpnWFFv dDV1cmFTdWFLV3VNc2lsYS1aakhLQkM3LW05MmZKUXB6bUI2WVJKM0xCZjBQUUFpdyQ8L2E+PGJy Pg0KPGJyPg0KSWYgeW91ciBtYWlsIGJvdW5jZXMgZnJvbSBDQ0wgd2l0aCA1LjcuMSBlcnJvciwg Y2hlY2s6PGJyPg0KJm5ic3A7Jm5ic3A7Jm5ic3A7Jm5ic3A7Jm5ic3A7IDxhIGhyZWY9Imh0dHBz Oi8vdXJsZGVmZW5zZS51cy92My9fX2h0dHA6Ly93d3cuY2NsLm5ldC9zcGFtbWVycy50eHRfXzsh IUcya3BNN3VNLVR6SUZjaHUhM3A5VGRTa1B2bFdZN3NwTEN6ZHJ3OEFTSGJVdmpqUWtacFNQSmdY UW90NXVyYVN1YUtXdU1zaWxhLVpqSEtCQzctbTkyZkpRcHptQjZZUkozTENQTkhCclV3JCIgaWQ9 Ik9XQTJhOTUwN2QxLTM1ODctNTQyNy0wMjNkLTY5MjlmNTQ1MmFhZiIgY2xhc3M9Ik9XQUF1dG9M aW5rIiBkYXRhLWF1dGg9Ik5vdEFwcGxpY2FibGUiPg0KaHR0cHM6Ly91cmxkZWZlbnNlLnVzL3Yz L19faHR0cDovL3d3dy5jY2wubmV0L3NwYW1tZXJzLnR4dF9fOyEhRzJrcE03dU0tVHpJRmNodSEz cDlUZFNrUHZsV1k3c3BMQ3pkcnc4QVNIYlV2ampRa1pwU1BKZ1hRb3Q1dXJhU3VhS1d1TXNpbGEt WmpIS0JDNy1tOTJmSlFwem1CNllSSjNMQ1BOSEJyVXckPC9hPjxicj4NCjxicj4NClJURkk6IDxh IGhyZWY9Imh0dHBzOi8vdXJsZGVmZW5zZS51cy92My9fX2h0dHA6Ly93d3cuY2NsLm5ldC9jaGVt aXN0cnkvYWJvdXRjY2wvaW5zdHJ1Y3Rpb25zL19fOyEhRzJrcE03dU0tVHpJRmNodSEzcDlUZFNr UHZsV1k3c3BMQ3pkcnc4QVNIYlV2ampRa1pwU1BKZ1hRb3Q1dXJhU3VhS1d1TXNpbGEtWmpIS0JD Ny1tOTJmSlFwem1CNllSSjNMRFpIck11dFEkIiBpZD0iT1dBZjkwMzFhNDYtOTNiMi1lMTg1LTM4 M2MtZTdiNmI4MjliNDE1IiBjbGFzcz0iT1dBQXV0b0xpbmsiIGRhdGEtYXV0aD0iTm90QXBwbGlj YWJsZSI+DQpodHRwczovL3VybGRlZmVuc2UudXMvdjMvX19odHRwOi8vd3d3LmNjbC5uZXQvY2hl bWlzdHJ5L2Fib3V0Y2NsL2luc3RydWN0aW9ucy9fXzshIUcya3BNN3VNLVR6SUZjaHUhM3A5VGRT a1B2bFdZN3NwTEN6ZHJ3OEFTSGJVdmpqUWtacFNQSmdYUW90NXVyYVN1YUtXdU1zaWxhLVpqSEtC QzctbTkyZkpRcHptQjZZUkozTERaSHJNdXRRJDwvYT48YnI+DQo8YnI+DQo8YnI+DQo8L2Rpdj4N CjwvYm9keT4NCjwvaHRtbD4NCg== --_000_BY3PR09MB81310C98B32822DDF5BAB344F6D62BY3PR09MB8131namp_-- From owner-chemistry@ccl.net Wed Jun 26 16:17:00 2024 From: "S M Bargeen Turzo smbargeen.turzo.2016###owu.edu" To: CCL Subject: CCL: AlphaFold 3 vs AutoDock Vina Message-Id: <-55178-240626135634-6267-q7bE1r+KtjrN1hYXltcRvg]-[server.ccl.net> X-Original-From: S M Bargeen Turzo Content-Type: multipart/alternative; boundary="0000000000002eee4f061bcebb67" Date: Wed, 26 Jun 2024 13:56:06 -0400 MIME-Version: 1.0 Sent to CCL by: S M Bargeen Turzo [smbargeen.turzo.2016+*+owu.edu] --0000000000002eee4f061bcebb67 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable Maybe relevant - Not sure how true to the original AF3 implementation this is but given that DeepMind decided to make this closed source there has been some efforts from the open source community to implement AF3. See github link below: https://github.com/lucidrains/alphafold3-pytorch On Wed, Jun 26, 2024 at 1:51=E2=80=AFPM Oleg Trott trott : caa.columbia.edu= < owner-chemistry*|*ccl.net> wrote: > > Sent to CCL by: Oleg Trott [trott:+:caa.columbia.edu] > Thanks! > > I hope someone re-runs DeepMind's calculations, with the same > settings, but using k-fold cross-validation (like what I did at > Columbia), where test and training sets are never related. This will > show how well (or poorly, as the case may be) the method generalizes > to unrelated proteins. > > > =3D=3D=3D=3D=3D > Oleg Trott, PhD > https://olegtrott.com > > > > > > On Tue, Jun 25, 2024 at 11:31=E2=80=AFAM Aydin Manzouri aydin.manzouri ~~ > gmail.com wrote: > > > > Truly insightful. Thanks a lot. > > > > On Tue, Jun 25, 2024 at 1:59=E2=80=AFAM Oleg Trott trott[-]caa.columbia= .edu > wrote: > >> > >> > >> Sent to CCL by: "Oleg Trott" [trott-$-caa.columbia.edu] > >> Hello, everyone! > >> > >> DeepMind's new AlphaFold 3 attempts to predict protein-ligand binding, > and > >> their publication compares it to AutoDock Vina (which I built). But > their > >> methodology seems strange. > >> > >> I wrote up my comments in a blog post. If you have an interest in AI > and/or > >> docking, I hope you'll find it insightful. > >> > >> https://olegtrott.substack.com/p/are-alphafolds-new-results-a-miracle>= > > E-mail to subscribers: CHEMISTRY*_*ccl.net or use:>> > >> E-mail to administrators: CHEMISTRY-REQUEST*_*ccl.net or use>> > >> > > > > -=3D This is automatically added to each message by the mailing script = =3D-> > > --0000000000002eee4f061bcebb67 Content-Type: text/html; charset="UTF-8" Content-Transfer-Encoding: quoted-printable
Maybe relevant - Not sure how true to the original AF3 implemen= tation this is but given that DeepMind decided to make this closed source t= here has been some efforts from the open source community to implement AF3.= See github link below:
https://github.com/lucidrains/alphafold3-pytorch
<= br>
On Wed,= Jun 26, 2024 at 1:51=E2=80=AFPM Oleg Trott trott : caa.columbia.edu <owner-chemistry*|*ccl.net> wrote:

Sent to CCL by: Oleg Trott [trott:+:caa.columbia.edu]
Thanks!

I hope someone re-runs DeepMind's calculations, with the same
settings, but using k-fold cross-validation (like what I did at
Columbia), where test and training sets are never related. This will
show how well (or poorly, as the case may be) the method generalizes
to unrelated proteins.


=3D=3D=3D=3D=3D
Oleg Trott, PhD
http= s://olegtrott.com





On Tue, Jun 25, 2024 at 11:31=E2=80=AFAM Aydin Manzouri aydin.manzouri ~~ gmail.com= <owner-chemistry_-_ccl.net> wrote:
>
> Truly insightful. Thanks a lot.
>
> On Tue, Jun 25, 2024 at 1:59=E2=80=AFAM Oleg Trott trott[-]caa.columbia.= edu <owner-chemistry*_*ccl.net> wrote:
>>
>>
>> Sent to CCL by: "Oleg=C2=A0 Trott" [trott-$-caa.columbia.e= du]
>> Hello, everyone!
>>
>> DeepMind's new AlphaFold 3 attempts to predict protein-ligand = binding, and
>> their publication compares it to AutoDock Vina (which I built). Bu= t their
>> methodology seems strange.
>>
>> I wrote up my comments in a blog post. If you have an interest in = AI and/or
>> docking, I hope you'll find it insightful.
>>
>> https://olegtrott.subs= tack.com/p/are-alphafolds-new-results-a-miracle>> E-mail to subsc= ribers: CHEMISTRY*_*ccl.net or use:>>
>> E-mail to administrators: CHEMISTRY-REQUEST*_*ccl.net or use>>
>>



-=3D This is automatically added to each message by the mailing script =3D-=
E-mail to subscribers: CHEMISTRY*|*ccl.net or use:
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/cgi-bin/ccl/s= end_ccl_message

E-mail to administrators: CHEMISTRY-REQUEST*|*ccl.net or use
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/cgi-bin/ccl/s= end_ccl_message
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/chemistry/sub_un= sub.shtml

Before posting, check wait time at: http://www.ccl.net

Job: http://www.ccl.net/jobs
Conferences: http://server.ccl.net/chemist= ry/announcements/conferences/

Search Messages: http://www.ccl.net/chemistry/sear= chccl/index.shtml
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/spammers.txt

RTFI: http://www.ccl.net/chemistry/aboutccl/ins= tructions/


--0000000000002eee4f061bcebb67-- From owner-chemistry@ccl.net Wed Jun 26 16:53:01 2024 From: "sharat chandra sharat.chndr(0)gmail.com" To: CCL Subject: CCL: AlphaFold 3 vs AutoDock Vina Message-Id: <-55179-240626135959-6749-/CNABjWP+3T+vqJhpMnUOQ|-|server.ccl.net> X-Original-From: sharat chandra Content-Type: multipart/alternative; boundary="000000000000bad594061bcec700" Date: Wed, 26 Jun 2024 13:59:35 -0400 MIME-Version: 1.0 Sent to CCL by: sharat chandra [sharat.chndr],[gmail.com] --000000000000bad594061bcec700 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable Hi Oleg, Thank you for sharing your thoughts. I found a similar vibe from another blog article. "https://www.inductive.bio/blog/strong-baseline-for-alphafold-3-docking" These folks (in the blog) are working on re-running Deepmind calculations. Best regards Sharat On Wed, Jun 26, 2024 at 1:35=E2=80=AFPM Oleg Trott trott : caa.columbia.edu= < owner-chemistry=ccl.net> wrote: > > Sent to CCL by: Oleg Trott [trott:+:caa.columbia.edu] > Thanks! > > I hope someone re-runs DeepMind's calculations, with the same > settings, but using k-fold cross-validation (like what I did at > Columbia), where test and training sets are never related. This will > show how well (or poorly, as the case may be) the method generalizes > to unrelated proteins. > > > =3D=3D=3D=3D=3D > Oleg Trott, PhD > https://olegtrott.com > > > > > > On Tue, Jun 25, 2024 at 11:31=E2=80=AFAM Aydin Manzouri aydin.manzouri ~~ > gmail.com wrote: > > > > Truly insightful. Thanks a lot. > > > > On Tue, Jun 25, 2024 at 1:59=E2=80=AFAM Oleg Trott trott[-]caa.columbia= .edu > wrote: > >> > >> > >> Sent to CCL by: "Oleg Trott" [trott-$-caa.columbia.edu] > >> Hello, everyone! > >> > >> DeepMind's new AlphaFold 3 attempts to predict protein-ligand binding, > and > >> their publication compares it to AutoDock Vina (which I built). But > their > >> methodology seems strange. > >> > >> I wrote up my comments in a blog post. If you have an interest in AI > and/or > >> docking, I hope you'll find it insightful. > >> > >> https://olegtrott.substack.com/p/are-alphafolds-new-results-a-miracle>= > > E-mail to subscribers: CHEMISTRY*_*ccl.net or use:>> > >> E-mail to administrators: CHEMISTRY-REQUEST*_*ccl.net or use>> > >> > > > > -=3D This is automatically added to each message by the mailing script = =3D-> > > --000000000000bad594061bcec700 Content-Type: text/html; charset="UTF-8" Content-Transfer-Encoding: quoted-printable
Hi Oleg,

T= hank you for sharing your thoughts.
I found a similar vibe fr= om another blog article.=C2=A0
"https://www.inductiv= e.bio/blog/strong-baseline-for-alphafold-3-docking" These folks (i= n the blog) are working on re-running Deepmind calculations.=C2=A0

Best regards
Sharat

On Wed, Jun 26, 2024 at= 1:35=E2=80=AFPM Oleg Trott trott : caa= .columbia.edu <owner-chem= istry=ccl.net> wrote:

Sent to CCL by: Oleg Trott [trott:+:caa.columbia.edu]
Thanks!

I hope someone re-runs DeepMind's calculations, with the same
settings, but using k-fold cross-validation (like what I did at
Columbia), where test and training sets are never related. This will
show how well (or poorly, as the case may be) the method generalizes
to unrelated proteins.


=3D=3D=3D=3D=3D
Oleg Trott, PhD
http= s://olegtrott.com





On Tue, Jun 25, 2024 at 11:31=E2=80=AFAM Aydin Manzouri aydin.manzouri ~~ gmail.com= <owner-chemistry_-_ccl.net> wrote:
>
> Truly insightful. Thanks a lot.
>
> On Tue, Jun 25, 2024 at 1:59=E2=80=AFAM Oleg Trott trott[-]caa.columbia.= edu <owner-chemistry*_*ccl.net> wrote:
>>
>>
>> Sent to CCL by: "Oleg=C2=A0 Trott" [trott-$-caa.columbia.e= du]
>> Hello, everyone!
>>
>> DeepMind's new AlphaFold 3 attempts to predict protein-ligand = binding, and
>> their publication compares it to AutoDock Vina (which I built). Bu= t their
>> methodology seems strange.
>>
>> I wrote up my comments in a blog post. If you have an interest in = AI and/or
>> docking, I hope you'll find it insightful.
>>
>> https://olegtrott.subs= tack.com/p/are-alphafolds-new-results-a-miracle>> E-mail to subsc= ribers: CHEMISTRY*_*ccl.net or use:>>
>> E-mail to administrators: CHEMISTRY-REQUEST*_*ccl.net or use>>
>>



-=3D This is automatically added to each message by the mailing script =3D-=
E-mail to subscribers: CHEMISTRY=ccl.net or use:
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/cgi-bin/ccl/s= end_ccl_message

E-mail to administrators: CHEMISTRY-REQUEST=ccl.net or use
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/cgi-bin/ccl/s= end_ccl_message
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/chemistry/sub_un= sub.shtml

Before posting, check wait time at: http://www.ccl.net

Job: http://www.ccl.net/jobs
Conferences: http://server.ccl.net/chemist= ry/announcements/conferences/

Search Messages: http://www.ccl.net/chemistry/sear= chccl/index.shtml
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/spammers.txt

RTFI: http://www.ccl.net/chemistry/aboutccl/ins= tructions/


--000000000000bad594061bcec700-- From owner-chemistry@ccl.net Wed Jun 26 17:28:01 2024 From: "Josiah Bjorgaard josiahbjorgaard[#]gmail.com" To: CCL Subject: CCL: AlphaFold 3 vs AutoDock Vina Message-Id: <-55180-240626145157-22703-dkKKCHwse7Zb1LubFNK4KQ,,server.ccl.net> X-Original-From: Josiah Bjorgaard Content-Type: multipart/alternative; boundary="0000000000003da29d061bcf8159" Date: Wed, 26 Jun 2024 11:51:29 -0700 MIME-Version: 1.0 Sent to CCL by: Josiah Bjorgaard [josiahbjorgaard,gmail.com] --0000000000003da29d061bcf8159 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable That was a great blog post on AF3 vs. Vina. The hype on AF3 is too strong. I'm curious if there are similar concerns with the RFDiffusion All Atom predictions? They claim a near-Vina quality result on PoseBusters and are similar in principle to the AF3 model. They further claim superior performance to Autodock Vina on a blind-test evaluation called CAMEO, which I'm not familiar with. They state the following in their paper ( https://www.biorxiv.org/content/10.1101/2023.10.09.561603v1.full): "To determine the extent to which the network is learning general principles of protein-small molecule interactions, we investigated the correlation between prediction accuracy and physically based correlates of protein-small molecule interaction affinity. We found that predictions for protein-small molecule complexes with high predicted affinity were more accurate than predictions for complexes predicted to bind weakly, further suggesting that RFAA is learning general principles rather than just memorizing global protein-small molecule interactions patterns" Curious what thoughts are on the validity of this approach - memorization or understanding? On Wed, Jun 26, 2024 at 11:23=E2=80=AFAM Oleg Trott trott : caa.columbia.ed= u < owner-chemistry(!)ccl.net> wrote: > > Sent to CCL by: Oleg Trott [trott:+:caa.columbia.edu] > Thanks! > > I hope someone re-runs DeepMind's calculations, with the same > settings, but using k-fold cross-validation (like what I did at > Columbia), where test and training sets are never related. This will > show how well (or poorly, as the case may be) the method generalizes > to unrelated proteins. > > > =3D=3D=3D=3D=3D > Oleg Trott, PhD > https://olegtrott.com > > > > > > On Tue, Jun 25, 2024 at 11:31=E2=80=AFAM Aydin Manzouri aydin.manzouri ~~ > gmail.com wrote: > > > > Truly insightful. Thanks a lot. > > > > On Tue, Jun 25, 2024 at 1:59=E2=80=AFAM Oleg Trott trott[-]caa.columbia= .edu > wrote: > >> > >> > >> Sent to CCL by: "Oleg Trott" [trott-$-caa.columbia.edu] > >> Hello, everyone! > >> > >> DeepMind's new AlphaFold 3 attempts to predict protein-ligand binding, > and > >> their publication compares it to AutoDock Vina (which I built). But > their > >> methodology seems strange. > >> > >> I wrote up my comments in a blog post. If you have an interest in AI > and/or > >> docking, I hope you'll find it insightful. > >> > >> https://olegtrott.substack.com/p/are-alphafolds-new-results-a-miracle>= > > E-mail to subscribers: CHEMISTRY*_*ccl.net or use:>> > >> E-mail to administrators: CHEMISTRY-REQUEST*_*ccl.net or use>> > >> > > > > -=3D This is automatically added to each message by the mailing script = =3D-> > > --0000000000003da29d061bcf8159 Content-Type: text/html; charset="UTF-8" Content-Transfer-Encoding: quoted-printable
That was a great blog post on AF3 vs. Vina. The hype on AF= 3 is too strong.
=C2=A0
I'm curious if there are similar concer= ns with the RFDiffusion All Atom predictions? They claim a near-Vina qualit= y result on PoseBusters and are similar in principle to the AF3 model. They= further claim superior performance to Autodock Vina on a blind-test evalua= tion called CAMEO, which I'm not familiar with.

They state the f= ollowing in their paper (https://www.biorxiv.org/content/10.1101/2023.10.= 09.561603v1.full):

"To determine the extent to which the network is learning general p= rinciples of protein-small molecule interactions, we investigated the corre= lation between prediction accuracy and physically based correlates of prote= in-small molecule interaction affinity. We found that predictions for prote= in-small molecule complexes with high predicted affinity=C2=A0were more accurate than predictions for complexes predic= ted to bind weakly, further suggesting that = RFAA is learning general principles rather than just memorizing global prot= ein-small molecule interactions patterns"

Curious what thoughts= are on the validity of this approach - memorization or understanding?

On Wed, Jun 26, 2024 at 11:23=E2=80=AFAM Oleg Trott trott := caa.columbia.edu <owner-chemistry(!)ccl.net> wrote:

Sent to CCL by: Oleg Trott [trott:+:caa.columbia.edu]
Thanks!

I hope someone re-runs DeepMind's calculations, with the same
settings, but using k-fold cross-validation (like what I did at
Columbia), where test and training sets are never related. This will
show how well (or poorly, as the case may be) the method generalizes
to unrelated proteins.


=3D=3D=3D=3D=3D
Oleg Trott, PhD
http= s://olegtrott.com





On Tue, Jun 25, 2024 at 11:31=E2=80=AFAM Aydin Manzouri aydin.manzouri ~~ gmail.com= <owner-chemistry_-_ccl.net> wrote:
>
> Truly insightful. Thanks a lot.
>
> On Tue, Jun 25, 2024 at 1:59=E2=80=AFAM Oleg Trott trott[-]caa.columbia.= edu <owner-chemistry*_*ccl.net> wrote:
>>
>>
>> Sent to CCL by: "Oleg=C2=A0 Trott" [trott-$-caa.columbia.e= du]
>> Hello, everyone!
>>
>> DeepMind's new AlphaFold 3 attempts to predict protein-ligand = binding, and
>> their publication compares it to AutoDock Vina (which I built). Bu= t their
>> methodology seems strange.
>>
>> I wrote up my comments in a blog post. If you have an interest in = AI and/or
>> docking, I hope you'll find it insightful.
>>
>> https://olegtrott.subs= tack.com/p/are-alphafolds-new-results-a-miracle>> E-mail to subsc= ribers: CHEMISTRY*_*ccl.net or use:>>
>> E-mail to administrators: CHEMISTRY-REQUEST*_*ccl.net or use>>
>>



-=3D This is automatically added to each message by the mailing script =3D-=
E-mail to subscribers: CHEMISTRY(!)ccl.net or use:
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/cgi-bin/ccl/s= end_ccl_message

E-mail to administrators: CHEMISTRY-REQUEST(!)ccl.net or use
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/cgi-bin/ccl/s= end_ccl_message
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/chemistry/sub_un= sub.shtml

Before posting, check wait time at: http://www.ccl.net

Job: http://www.ccl.net/jobs
Conferences: http://server.ccl.net/chemist= ry/announcements/conferences/

Search Messages: http://www.ccl.net/chemistry/sear= chccl/index.shtml
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/spammers.txt

RTFI: http://www.ccl.net/chemistry/aboutccl/ins= tructions/


--0000000000003da29d061bcf8159-- From owner-chemistry@ccl.net Wed Jun 26 19:26:00 2024 From: "Oleg Trott trott#caa.columbia.edu" To: CCL Subject: CCL: AlphaFold 3 vs AutoDock Vina Message-Id: <-55181-240626165725-13963-dHGIIWkNRpFFEQ4u78l6bg%server.ccl.net> X-Original-From: Oleg Trott Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset="UTF-8" Date: Wed, 26 Jun 2024 13:56:52 -0700 MIME-Version: 1.0 Sent to CCL by: Oleg Trott [trott^^^caa.columbia.edu] Brian, What I'm suggesting is this: 1. Proteins in the test dataset should be grouped by sequence similarity (If you have compute to burn, you can skip this and basically put each protein in its own cluster) 2. For each cluster, training should be re-done using only proteins that are unrelated to the cluster. This will give us a "generalization accuracy" for AF3 (generalization to unrelated proteins). I'm not familiar with this dataset personally. But someone alerted me that it has many non-drug-like compounds that are common in the PDB, but not very relevant in drug discovery. Perhaps they should be excluded from the tests? Oleg ===== Oleg Trott, PhD https://olegtrott.com On Wed, Jun 26, 2024 at 1:26 PM Bennion, Brian bennion1]|[llnl.gov wrote: > > I am fairly new in this area so my apologies. > Will k-fold cross validation split datasets by similarity as well as identity? The example in your post while very helpful to understand your point is highly simplified. > > Would C and C' for instance be seperated into different sets as well as multiple instances of C? > > Scaffold splitting when done properly will ensure that highly similar (tanimoto scores) compounds are not found in training/validation/test sets. > > Thanks for any help understanding this aspect of the conversation. > > brian > > > ________________________________ > From: owner-chemistry+bennion1==llnl.gov**ccl.net on behalf of Oleg Trott trott : caa.columbia.edu > Sent: Wednesday, June 26, 2024 7:52 AM > To: Bennion, Brian > Subject: CCL: AlphaFold 3 vs AutoDock Vina > > > Sent to CCL by: Oleg Trott [trott:+:caa.columbia.edu] > Thanks! > > I hope someone re-runs DeepMind's calculations, with the same > settings, but using k-fold cross-validation (like what I did at > Columbia), where test and training sets are never related. This will > show how well (or poorly, as the case may be) the method generalizes > to unrelated proteins. > > > ===== > Oleg Trott, PhD > https://urldefense.us/v3/__https://olegtrott.com__;!!G2kpM7uM-TzIFchu!3p9TdSkPvlWY7spLCzdrw8ASHbUvjjQkZpSPJgXQot5uraSuaKWuMsila-ZjHKBC7-m92fJQpzmB6YRJ3LAZt4GZGA$ > > > > > > On Tue, Jun 25, 2024 at 11:31 AM Aydin Manzouri aydin.manzouri ~~ > gmail.com wrote: > > > > Truly insightful. Thanks a lot. > > > > On Tue, Jun 25, 2024 at 1:59 AM Oleg Trott trott[-]caa.columbia.edu wrote: > >> > >> > >> Sent to CCL by: "Oleg Trott" [trott-$-caa.columbia.edu] > >> Hello, everyone! > >> > >> DeepMind's new AlphaFold 3 attempts to predict protein-ligand binding, and > >> their publication compares it to AutoDock Vina (which I built). But their > >> methodology seems strange. > >> > >> I wrote up my comments in a blog post. If you have an interest in AI and/or > >> docking, I hope you'll find it insightful. > >> > >> https://urldefense.us/v3/__https://olegtrott.substack.com/p/are-alphafolds-new-results-a-miracle__;!!G2kpM7uM-TzIFchu!3p9TdSkPvlWY7spLCzdrw8ASHbUvjjQkZpSPJgXQot5uraSuaKWuMsila-ZjHKBC7-m92fJQpzmB6YRJ3LAhjWUW2Q$ >> E-mail to subscribers: CHEMISTRY*_*ccl.net or use:>> > >> E-mail to administrators: CHEMISTRY-REQUEST*_*ccl.net or use> https://urldefense.us/v3/__http://www.ccl.net/cgi-bin/ccl/send_ccl_message__;!!G2kpM7uM-TzIFchu!3p9TdSkPvlWY7spLCzdrw8ASHbUvjjQkZpSPJgXQot5uraSuaKWuMsila-ZjHKBC7-m92fJQpzmB6YRJ3LDKQAR6UA$> https://urldefense.us/v3/__http://www.ccl.net/cgi-bin/ccl/send_ccl_message__;!!G2kpM7uM-TzIFchu!3p9TdSkPvlWY7spLCzdrw8ASHbUvjjQkZpSPJgXQot5uraSuaKWuMsila-ZjHKBC7-m92fJQpzmB6YRJ3LDKQAR6UA$> https://urldefense.us/v3/__http://www.ccl.net/chemistry/sub_unsub.shtml__;!!G2kpM7uM-TzIFchu!3p9TdSkPvlWY7spLCzdrw8ASHbUvjjQkZpSPJgXQot5uraSuaKWuMsila-ZjHKBC7-m92fJQpzmB6YRJ3LCPWW2CVw$ > > Before posting, check wait time at: https://urldefense.us/v3/__http://www.ccl.net__;!!G2kpM7uM-TzIFchu!3p9TdSkPvlWY7spLCzdrw8ASHbUvjjQkZpSPJgXQot5uraSuaKWuMsila-ZjHKBC7-m92fJQpzmB6YRJ3LDlpU830Q$ > > Job: https://urldefense.us/v3/__http://www.ccl.net/jobs__;!!G2kpM7uM-TzIFchu!3p9TdSkPvlWY7spLCzdrw8ASHbUvjjQkZpSPJgXQot5uraSuaKWuMsila-ZjHKBC7-m92fJQpzmB6YRJ3LDCTKWjAQ$ > Conferences: https://urldefense.us/v3/__http://server.ccl.net/chemistry/announcements/conferences/__;!!G2kpM7uM-TzIFchu!3p9TdSkPvlWY7spLCzdrw8ASHbUvjjQkZpSPJgXQot5uraSuaKWuMsila-ZjHKBC7-m92fJQpzmB6YRJ3LBXlUqlyg$ > > Search Messages: https://urldefense.us/v3/__http://www.ccl.net/chemistry/searchccl/index.shtml__;!!G2kpM7uM-TzIFchu!3p9TdSkPvlWY7spLCzdrw8ASHbUvjjQkZpSPJgXQot5uraSuaKWuMsila-ZjHKBC7-m92fJQpzmB6YRJ3LBf0PQAiw$> https://urldefense.us/v3/__http://www.ccl.net/spammers.txt__;!!G2kpM7uM-TzIFchu!3p9TdSkPvlWY7spLCzdrw8ASHbUvjjQkZpSPJgXQot5uraSuaKWuMsila-ZjHKBC7-m92fJQpzmB6YRJ3LCPNHBrUw$ > > RTFI: https://urldefense.us/v3/__http://www.ccl.net/chemistry/aboutccl/instructions/__;!!G2kpM7uM-TzIFchu!3p9TdSkPvlWY7spLCzdrw8ASHbUvjjQkZpSPJgXQot5uraSuaKWuMsila-ZjHKBC7-m92fJQpzmB6YRJ3LDZHrMutQ$ > > From owner-chemistry@ccl.net Wed Jun 26 20:01:00 2024 From: "Josiah Bjorgaard josiahbjorgaard^^gmail.com" To: CCL Subject: CCL: AlphaFold 3 vs AutoDock Vina Message-Id: <-55182-240626190706-24708-zjtMrI89pQ57NjMUc1vGmg_-_server.ccl.net> X-Original-From: Josiah Bjorgaard Content-Type: multipart/alternative; boundary="000000000000395c97061bd31256" Date: Wed, 26 Jun 2024 16:06:46 -0700 MIME-Version: 1.0 Sent to CCL by: Josiah Bjorgaard [josiahbjorgaard%a%gmail.com] --000000000000395c97061bd31256 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable Maybe important to note, although not AF3, there are similar models which are actually open source - DiffDock and recently DynamicBind https://www.nature.com/articles/s41467-024-45461-2 -- DynamicBind DynamicBind claims significant improvements over Vina, Gnina, and Glide, but seem to also say that DiffDock is also better, which is not the case in other benchmarks. On Wed, Jun 26, 2024 at 3:42=E2=80=AFPM sharat chandra sharat.chndr(0)gmail= .com < owner-chemistry#%#ccl.net> wrote: > Hi Oleg, > > Thank you for sharing your thoughts. > I found a similar vibe from another blog article. > "https://www.inductive.bio/blog/strong-baseline-for-alphafold-3-docking" > These folks (in the blog) are working on re-running Deepmind calculations= . > > Best regards > Sharat > > On Wed, Jun 26, 2024 at 1:35=E2=80=AFPM Oleg Trott trott : caa.columbia.e= du < > owner-chemistry^^^ccl.net> wrote: > >> >> Sent to CCL by: Oleg Trott [trott:+:caa.columbia.edu] >> Thanks! >> >> I hope someone re-runs DeepMind's calculations, with the same >> settings, but using k-fold cross-validation (like what I did at >> Columbia), where test and training sets are never related. This will >> show how well (or poorly, as the case may be) the method generalizes >> to unrelated proteins. >> >> >> =3D=3D=3D=3D=3D >> Oleg Trott, PhD >> https://olegtrott.com >> >> >> >> >> >> On Tue, Jun 25, 2024 at 11:31=E2=80=AFAM Aydin Manzouri aydin.manzouri ~= ~ >> gmail.com wrote: >> > >> > Truly insightful. Thanks a lot. >> > >> > On Tue, Jun 25, 2024 at 1:59=E2=80=AFAM Oleg Trott trott[-]caa.columbi= a.edu >> wrote: >> >> >> >> >> >> Sent to CCL by: "Oleg Trott" [trott-$-caa.columbia.edu] >> >> Hello, everyone! >> >> >> >> DeepMind's new AlphaFold 3 attempts to predict protein-ligand binding= , >> and >> >> their publication compares it to AutoDock Vina (which I built). But >> their >> >> methodology seems strange. >> >> >> >> I wrote up my comments in a blog post. If you have an interest in AI >> and/or >> >> docking, I hope you'll find it insightful. >> >> >> >> https://olegtrott.substack.com/p/are-alphafolds-new-results-a-miracle= >> >> E-mail to subscribers: CHEMISTRY*_*ccl.net or use:>> >> >> E-mail to administrators: CHEMISTRY-REQUEST*_*ccl.net or use>> >> >> >> >> >> >> -=3D This is automatically added to each message by the mailing script = =3D- >> E-mail to subscribers: CHEMISTRY^^^ccl.net or use:>> >> E-mail to administrators: CHEMISTRY-REQUEST^^^ccl.net or use>> >> >> --000000000000395c97061bd31256 Content-Type: text/html; charset="UTF-8" Content-Transfer-Encoding: quoted-printable
Maybe important to note, although not AF3, there are = similar models which are actually open source - DiffDock and recently Dynam= icBind

https://www.nature.com/articles/s41467-024-45461-2 = -- DynamicBind

DynamicBind claims significant improvements over Vina= , Gnina, and Glide, but seem to also say that DiffDock is also better, whic= h is not the case in other benchmarks.


On Wed, Jun 26, = 2024 at 3:42=E2=80=AFPM sharat chandra sharat.chndr(0)gmail.com <owner-= chemistry#%#ccl.net> wrote:
Hi Oleg,

Thank you for sharing your thoughts.
I foun= d a similar vibe from another blog article.=C2=A0
"https://www.inductive.bio/blog/strong-baseline-for-alpha= fold-3-docking" These folks (in the blog) are working on re-runnin= g Deepmind calculations.=C2=A0

Best regards
<= div>Sharat

On Wed, Jun 26, 2024 at 1:35=E2=80=AFPM Oleg Trott trott : = caa.columbia.edu = <ow= ner-chemistry^^^ccl.net> wrote:

Sent to CCL by: Oleg Trott [trott:+:caa.columbia.edu]
Thanks!

I hope someone re-runs DeepMind's calculations, with the same
settings, but using k-fold cross-validation (like what I did at
Columbia), where test and training sets are never related. This will
show how well (or poorly, as the case may be) the method generalizes
to unrelated proteins.


=3D=3D=3D=3D=3D
Oleg Trott, PhD
http= s://olegtrott.com





On Tue, Jun 25, 2024 at 11:31=E2=80=AFAM Aydin Manzouri aydin.manzouri ~~ gmail.com= <owner-chemistry_-_ccl.net> wrote:
>
> Truly insightful. Thanks a lot.
>
> On Tue, Jun 25, 2024 at 1:59=E2=80=AFAM Oleg Trott trott[-]caa.columbia.= edu <owner-chemistry*_*ccl.net> wrote:
>>
>>
>> Sent to CCL by: "Oleg=C2=A0 Trott" [trott-$-caa.columbia.e= du]
>> Hello, everyone!
>>
>> DeepMind's new AlphaFold 3 attempts to predict protein-ligand = binding, and
>> their publication compares it to AutoDock Vina (which I built). Bu= t their
>> methodology seems strange.
>>
>> I wrote up my comments in a blog post. If you have an interest in = AI and/or
>> docking, I hope you'll find it insightful.
>>
>> https://olegtrott.subs= tack.com/p/are-alphafolds-new-results-a-miracle>> E-mail to subsc= ribers: CHEMISTRY*_*ccl.net or use:>>
>> E-mail to administrators: CHEMISTRY-REQUEST*_*ccl.net or use>>
>>



-=3D This is automatically added to each message by the mailing script =3D-=
E-mail to subscribers: CHEMISTRY^^^ccl.net or use:
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/cgi-bin/ccl/s= end_ccl_message

E-mail to administrators: CHEMISTRY-REQUEST^^^ccl.net or use
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/cgi-bin/ccl/s= end_ccl_message
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/chemistry/sub_un= sub.shtml

Before posting, check wait time at: http://www.ccl.net

Job: http://www.ccl.net/jobs
Conferences: http://server.ccl.net/chemist= ry/announcements/conferences/

Search Messages: http://www.ccl.net/chemistry/sear= chccl/index.shtml
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/spammers.txt

RTFI: http://www.ccl.net/chemistry/aboutccl/ins= tructions/


--000000000000395c97061bd31256--