CASP13

Timetable
April 2018 - Start of the registration for CASP13 prediction experiment.
April 18, 2018 - Start of the testing of server connectivity ("dry run" for server predictors).
May 1, 2018 - Release of the first CASP13 modeling targets.
May/June 2018 - Early bird registration for the December CASP13 conference.
July 16, 2018 - Last date for releasing regular targets.
July 31, 2018 - End of the regular modeling season.
August 17, 2018 - End of the refinement and data-assisted modeling season.
September 2018 - Collection of abstracts describing the methods used in CASP13.
October/November 2018 - Invitations to groups with the most accurate models and the most interesting methods to give talks at the CASP13 conference.
Novermber 2018 - Program of the conference finalized.
December 2018 - CASP13 Conference.

Rosetta is ready to partecipate??
During the others CASP, R@H released wus with "casp" in their name

---

CASP13

Rosetta is ready to partecipate??
During the others CASP, R@H released wus with "casp" in their name

uh, I had not seen this line at the bottom of the home page

Scientific advisory board
David Baker, University of Washington
Michael Feig, Michigan State University
Nick Grishin, University of Texas
Andrzej Joachimiak, Argonne National Lab
David Jones, University College, London
Rachel Karchin, John Hopkins University
Chaok Seok, Seoul National University
Michael Sternberg, Imperial College, London

---

We have one more week to help bring your servers in contact with our distribution server and make sure your submission format is correct. The first CASP13 target will be released as planned, on May 1.

---

We are set to start the 13th CASP prediction season. As of today, more than 160 groups have registered, including 80 servers. Dry runs for servers are now over, and if you did not manage to test your server during those runs we will do our best to help you in the first days of CASP13. If you are planning to participate (either as a human expert group or a server group), but have not registered yet - please do this as soon as possible.

We will start CASP13 tomorrow with one relatively easy for modeling target, T0949.

On Wednesday, we will release a 'burning' target, T0950. Crystallographers who solved this target want to make it publicly available without any delays and expect the structure to be released by the PDB in the closest week or two. This is a difficult modeling target and we don't want to loose it for CASP. Therefore, this target will be released for prediction with a shorter 2-week deadline. If we learn that the prediction window for this target can be extended (or vise versa need to be further shortened) we will inform you immediately. If you participate as an expert group - please make this target your first week priority.

Two more targets, a server-only target and a human target representing short dimerization domain, will be released on Thursday.

On Friday we will suggest for modeling a difficult heteromeric structure (A3B1 stoichiometry), which will be released as 3 separate targets: a full complex target H0953, and two constitutive subunits (T0953s1 and T0953s2). For those submitting models for the whole complex there is no need to additionally send models on subunit targets (please see our format page for details http://predictioncenter.org/casp13/index.cgi?page=format#TS).

---

Dear CASP13 participants,

Web server problems:
As most of you are aware by now, we had a power outage in our server room on Friday night (around 23:30 Pacific Time), which knocked down both CASP web servers (the main one and the backup). The passed weekend and this Monday turned to be quite stressful for the Prediction Center team, but we managed to restore the service back to normal by now.

Adjusted target deadlines:
Since servers and QA groups could not send us predictions according to the originally scheduled deadlines, we have adjusted submission schedules for a number of targets. Please consult our Target List page for the new deadlines and make sure we received your predictions by then. Sorry for the inconvenience this is causing you.

New targets:
We will continue releasing targets tomorrow, Tuesday May 15, with the target T0963. Targets T0961 and T0962 that were scheduled for this Monday's release will be released next Monday, May 21. Four more targets will be released on Wednesday and Thursday, including first two targets selected for CAPRI. Friday will bring for prediction a heteromeric complex and its subunits.

---

Week 3 summary:
* >150 groups sending predictions
* 23 regular (single-sequence) targets
* 3 heteromeric targets
* 2 CAPRI nominated targets
* >6,000 predictions accepted

New targets:
This week we will release one heteromeric target (on Friday) and 9 single-sequence targets (including two subunits of the Friday multimer). Three of these targets are selected for CAPRI experiment. Those participating in SAXS-assissted modeling should expect the first target in this category on Tuesday. Also, depending on the results of preliminary evaluation of T0949, we may release this target for the refinement experiment later in the week.

---

Plans for week 5 (May 28 - June 1) (2018-5-25)
There will be no targets on Monday.

Tuesday through Friday we will release:
* 7 regular targets, including 1 CAPRI
* 1 heteromeric target (Friday)
* 2 SAXS-assisted targets
* at least 1 refinement target

---

Target release:
* 7 regular targets, including 2 CAPRI (Tue, Wed)
Warning. Server groups should be prepared for processing large targets on Wednesday (target T0984, ~750 res.) and Thursday (target T0985, 850 res.)
* 1 heteromeric target (Fri)
* 2+ refinement targets

H0968:
The original stoichiometry for this target was suggested as A1B1. The assembly assessors think that higher-agglomeration arrangement is also feasible, so we are changing the suggested stoichiometry from A1B1 to A2B2 so that those wishing to send us models as an A2B2 complex could do that.

Summary for the first 5 weeks of CASP:
* 37 tertiary structure prediction targets
* 5 heteromer targets
* 2 refinement targets
* 3 assisted prediction targets (7 with subunits)
* 14,000 predictions from 164 groups

---

We are wondering if any CASP participants would like us to organize a "Data Guided Prediction" question and answer session with the people providing SAXS, Cross-link, and/or NMR data. Please respond to casp@predictioncenter.org so we have an idea of how much need there is for such a Q&A session.

---

Target release plans for week 7:
* 8 regular targets
* 1 heteromeric target (Fri)
* 1 SAXS-assisted target
* 2 XL-MS assisted targets
* ? refinement targets
* 2 CAPRI targets (Tue, Fri)

T0970:
Analyzing structure of T0970 we noticed that the original sequence was provided to CASP with trimmed off disorder regions. The correct sequence is:
KRSGFLTLGYRGSY[TTVRDNQADAKFRR]VARIMVCGRIALAKEVFGDTLNESRD[PDRQP]EKYTSRFYLKFTYLEQAFDRLSEAGFHMVACNSSGTAAF[VNQ]YRDDKIWSSYTEYIFFRP
(three disordered regions are provided in brackets).
Even though the absence of disorder residues in models will not affect the evaluation (we would cut them off for evaluation purposes anyways), we do realize that not knowing about the presence of these residues in the protein sequence might distort the modeling procedure as prediction methods could try to connect residues Y14 and V15 in the provided sequence by a peptide bond, while in fact these residues are 14 residues apart in the complete protein sequence. Please check your predictions and adjust structures if needed for this sequence mistake (please continue using residue numbering from the originally released sequence).

Contact predictions for long targets:
CASP system allows accepting predictions up to 300,000 lines in a file. For long targets, this may not be enough to submit the whole contact map. However, in practice submitting the whole contact map is rarely needed.
First, you can eliminate pairs of residues separated by fewer than 12 positions as short-range contacts have never been assessed in CASP contact prediction. Next, you don't need to include contact pairs predicted with close to zero probability as missing pairs are assumed to be predicted with 0 probability anyways. Following these two advises can substantially reduce your file size. If you nevertheless think that the whole contact map may be beneficial for evaluation, please submit the reduced prediction through the CASP gateway and, in addition to this, write to us and we would instruct you where to upload your complete contact map file.

---

T0970:
Analyzing structure of T0970 we noticed that the original sequence was provided to CASP with trimmed off disorder regions. The correct sequence is:
KRSGFLTLGYRGSY[TTVRDNQADAKFRR]VARIMVCGRIALAKEVFGDTLNESRD[PDRQP]EKYTSRFYLKFTYLEQAFDRLSEAGFHMVACNSSGTAAF[VNQ]YRDDKIWSSYTEYIFFRP
(three disordered regions are provided in brackets).
Even though the absence of disorder residues in models will not affect the evaluation (we would cut them off for evaluation purposes anyways), we do realize that not knowing about the presence of these residues in the protein sequence might distort the modeling procedure as prediction methods could try to connect residues Y14 and V15 in the provided sequence by a peptide bond, while in fact these residues are 14 residues apart in the complete protein sequence. Please check your predictions and adjust structures if needed for this sequence mistake (please continue using residue numbering from the originally released sequence).

Some of these wus has gone on Ralph

---

Target release plans for week 8:
* 6 regular targets, including 3 targets larger than 500 residues
* 1 SAXS-assisted target (Friday)
* some refinement targets (depending on the evaluation results)
* 2 or 3 CAPRI targets

Data-assisted targets:
This week we released first two cross-linking assisted targets - H0953 and H0957. More to follow in the last week of June.
In addition to the four already released SAXS-assisted targets, we are expecting at least five more later in the prediction season.
First sparse NMR data-assisted targets are expected in the end of this month.

T0999 (Friday):
Next week we are scheduled to reach another milestone in the CASP history: Friday will mark the last target from the first thousand of CASP targets overall. And it will be a special target! T0999 will be the largest target ever released in CASP. It is a homodimer of ~3200 residues total (~1600 residues in each chain). Predictors should be able to easily find that each chain of this target is a fusion of 5 enzymes. Even though prediction of individual domains is not expected to be very challenging (all five domains have multiple good templates), the arrangement of 10 domains within the complex is unknown and may be rather a difficult task! To make the story even more interesting, we have the SAXS and cross-linking data for this target, and planning to release those in turn after the regular prediction is over. It would be interesting to see if these data help generate better quaternary structure for such a large target.

---

Target release plans for week 9:
* 6 regular targets, including one longer target on Mon, Wed-Fri, and two shorter targets on Tuesday
* 2 SAXS-assisted targets
* 1 XL-MS -assisted target
* some refinement targets (depending on the evaluation results)
* no CAPRI targets

Please note that the next week we are starting second thousand of regular CASP targets. A request to server predictors: please make sure your scripts can process regular targets with the second symbol in the target name not being 0 any more (e.g., T1000).

---

Next week will be the second to last in a typical 11-week CASP target release season. We plan to release:
* 6 regular targets (no targets on Wednesday)
* 2 SAXS -assisted targets (Tue and Fri)
* 1 XL-MS -assisted target (same target H0957, different X-linking data acquisition technique - please comment in REMARKS on the use of a particular XLMS data set(s) )
* 5-6 refinement targets
* 1 CAPRI target
* 1 SANS -assisted target (a pilot, non-competitive trial)
* 2 simulated NMR data-assisted targets

T0988: canceled
It has been brought to our attention that target T0988 has virtually identical sequence to a previous CASP12 target T0881, and therefore we are canceling it.

---

arget release plan for July 9-13:
Next week is the last full week of target release in CASP13. We plan to go with 2 new targets per day Monday through Friday, including 3 heterocomplexes on Tue-Thu. We would also have 1 SAXS -assisted target (most likely the last SAXS -assisted target this season), 2+ refinement targets, and 1 cross-linking target (depending on the experimental data availability).

Week of July 16:
According to the announced CASP13 timetable, July 16 was planned as the last day for releasing regular targets. As of today, we do not have any targets for that day, but we do have a couple of good leads that may result in a few targets. If we get more targets than we can fit in a one-day release, we may extend the target release season by one or two days in order to accommodate those targets in CASP13. We will announce the final plans not later than Thursday of the next week.

Refinement and data-assisted targets:
We will continue releasing refinement and data-assisted targets through the first week of August.

Meeting registration:
We are planning to open registration for CASP13 conference next week.

Summary for the first 10 weeks of CASP:
* 71 tertiary structure prediction targets
* 7 heteromer targets
* 12 refinement targets
* 10 SAXS data sets resulting in 16 prediction targets
* 3+1 XL-MS data sets resulting in 12 prediction targets
* ~35,000 predictions from 179 groups

---

This week we were able to secure 3 very interesting cryo-EM targets. To accommodate them all in CASP, we will need one more day beyond what was originally planned. The targets will be released one per day, tomorrow, next Monday and next Tuesday, July 17. This will conclude the CASP13 regular target release season!

We received a request not to shorten prediction period for the last targets, so that enough time is allotted for their prediction. To accommodate it, we will set prediction deadlines for the last three targets in the first week of August.

---

Today (July 17) we released the last regular CASP13 target. All in all, we released 90 tertiary structure prediction targets, quite a representative data set ! The CASP13 target set differs from those of previous CASPs in several aspects. First, we had more than ever hetero-complex targets - 13. Second, we had many more large targets: 22 targets longer than 500 residues in CASP13 compared to 6 or less in five previous CASPs. Third, we had 8 targets (13 different subunits) solved with the cryo-EM technique, compared to 3 subunits in CASP12 and none before that. Finally, we had many more targets for assembly prediction, what also resulted in more targets for CAPRI (20 total in CASP13).

===
Refinement targets:
We will continue releasing refinement targets throughout the next 3 weeks. This week we will release three targets, one per day on Wed, Thu and Fri.

===
Data-assisted targets:

SAXS-assisted: we have released all SAXS-assisted targets (11 data sets resulting in 17 targets). Thanks to Greg Hura and Susan Tsutakawa (J.Tainer's group, LBL) for preparing experimental data on these targets.

SANS-assisted: we have released all SANS-assisted targets (1). Thanks to Anne Martel and Sergei Grudinin for preparing the data.

X-linking- assisted: we have released 5 XL-MS data sets (Alexander Leitner's group) and 2 XL-MS data sets (Juri Rappsilber's group) resulting in 17 targets. Three more data sets from A. Leitner's group will be released between now and July 30. More data sets from J. Rappsilber's group are possible.

NMR-simulated and real NMR-assisted: over the next 3 weeks, we will release contacts derived from NMR spectra that have been simulated for several FM targets (Gaetano Montelione's group). In addition, a real NMR data set has been generated for one CASP target, and a second real NMR data set for another protein is anticipated. Predictors may use these NMR data, together with contact predictions (ECs), crosslink data, and/or SAXS data, for data-guided structure prediction. A description of these sparse NMR data will be provided with the data. The first target is expected the coming Friday, July 20.

---

End of target release in CASP13 (2018-8-10)
Today we have finished releasing targets in CASP13. All in all, we released 90 regular targets; 13 hetero-multimer targets; 31 refinement targets; and 49 assisted modeling prediction targets in 5 data-assisted categories - SAXS, XL-MS, NMR (simulated and real data), SANS and FRET.

This past week we had quite an intense schedule for releasing refinement and data-assisted targets: 9 new refinement targets; 2 XL-MS targets; 9 NMR -assisted targets (8 simulated datasets and 1 real dataset); and 1 FRET data-assisted modeling target (F0964).

FRET data type is new in CASP. Similarly to SANS-assisted results, the FRET-assisted results will not be competitively assessed in CASP13. Rather, with this target we are initiating a discussion and allowing for a learning experience with the development of future FRET-based modeling techniques in mind. We are providing ample time to model this target with the closing date of September 30 (pending approval from the structure's author). The molecule’s two domains seem open to homology modeling but the relative domain orientation is not known (we have the structure of only the second of the two domains starting with the residues VIVIGDE). The FRET data suggest that there may be a distribution of domain orientations, opening a challenge of dynamic modeling in CASP. This topic will be discussed at the CASP13 meeting.

The description of the FRET data and an overview of the results for target F0964 can be found at http://predictioncenter.org/casp13/doc/FRET_CASP13_AGSeidel.pdf.
The detailed FRET data will be released progressively as they become available.

---

We have opened the webpage for collecting methods abstracts. The deadline for submitting the abstracts is September 22, 2018. Please read the instructions carefully and address all the questions specified in the Abstract submission form (http://predictioncenter.org/casp13/abstracts.cgi). Please remember that, as usually, your contributions will be taken into account in choosing presentations for the meeting.

---

CASP13 featured 90 protein targets. Over 57,000 predictions were submitted, including close to 36,000 tertiary structure models, 3,000 oligomeric models, 2,000 data-assisted models, 4,000 sets of contact predictions, 4,000 refined models, and 9,000 sets of accuracy estimations.

The conference will feature assessment talks, reports from the best predictors, and keynote presentations. The keynote speakers represent two current areas of special interest in CASP, machine learning /deep learning in structural biology and cryo-EM driven modeling:
* David Koes (U Pittsburgh), is a pioneer in the application of convolutional neural networks to molecular docking
* Wah Chiu (Stanford), is a leader in cryo-electron microscopy.

CASP13 will be held December 1-4, 2018, at the Iberostar Paraiso Maya resort (site of the CASP11 conference in 2014, destination airport is Cancun - CUN).

---