The new app 5.01 has some new science stuff and some new error-catching stuff that we're testing.

On science, we're testing a more aggressive minimization protocol (it led to a lot of errors in ralph release 4.98; we think we've got the fix), and a cool folding procedure that involves "stapling" parts of the backbone together and making cuts elsewhere. The latter procedure is called "jumping" for esoteric reasons.

On error-catching, we now have the app aborts if it hasn't made progress after 5 restarts (this should prevent clients with big jobs from tackling the same hard workunits over and over again).

---

We're beginning to hear that some machines are having difficulties
with the following jobs and the new app 5.01:

FACONTACTS_NOFILTERS_...
HBLR_1.0_1n0u_ROT_TRIALS_TRIE

If you suspect that your machine is taking too
long to complete the workunits above, please abort them!
We've cancelled the sendout of those workunits, and
are investigating the problem.

---

A quick update for all of you who have been paying attention
to our efforts on Ralph (5.01-5.06). The watchdog timer has
taken unexpectedly long to develop, due to the large
number of ways that clients are running (saving in memory;
frequency of switching apps; etc.) -- we're thankful
that we can test our apps so rapidly over there! We think
we've got a winner, but won't know until tomorrow. In addition
to the watchdog timer, we've added and tested some very
cool new modes in Rosetta, so you can expect some new
stuff on your screens soon!

---

Rosetta 5.06 is up. This version has some new error-catching features and some new science:

(1) Multiple checkpoints. Rosetta saves more information along
the way. If its preempted and restarted, now it doesn't have to
start from scratch. For large jobs, these checkpoints have tripled
the rate of structure prediction on our test project RALPH!

(2) Watchdog timer. This timer will abort jobs that appear
to be stuck in infinite loops (no progress for an hour) or have
gone over the preferred CPU run time by four-fold. This should
help those clients that cannot be continuously monitored,
like the ones from big computer farms.

(3) "Jumping" We're also implementing a new simulation protocol
that works better for large complex proteins, with some
parts of the chain glue together and other parts broken to
release the strain. You'll see some pretty interesting simulations!

---

Rosetta 5.07 is up. This version has a minor fix for the cpu run time user preference. More information about the bug that was fixed can be found at this tread.

Rosetta 5.12 is up!

1. We've centered and rescaled the graphics for large proteins so that they don't spend a lot of the time out of the box.

2. We've made it possible to send out workunit descriptions with the graphics. These won't be sent out with previously queued workunits, but we hope to at least have some text for new workunits throughout the summer!

3. We made some science fixes that will allow us to combine the "jumping" (strand-break) protocol with full-atom refinement.

---

A couple more points about Rosetta@home 5.12:

(1) We're on a mission to cut down memory requirements for Rosetta@home. You'll hopefully see a smaller memory footprint with 5.12 compared to 5.07, and it may even get smaller on the next application update!

(2) Some users are noticing a slowdown in other applications, such as Windows Media Player, when using the latest Rosetta@home. If you are seeing something similar, please post the workunit and your type of computer here, so that we can track it down.

---

Rosetta 5.13 has the same rosetta code as 5.12 but we've removed
some of the Boinc debugging code that was causing slowdowns on
a fraction of Windows machines. See this thread for more info on
what was happening, and how we fixed it.

---

Rosetta 5.16:

(1) We're continuing our efforts to reduce memory usage by typical workunits by rosetta@home. You can expect an even further reduction in memory footprint in our next update.

(2) We're testing a new science mode which uses the sequence and structural information from homologous proteins in an early phase of the simulation, but then returns to the target protein sequence in the final refinement phase. This mode appears to have a larger memory footprint than typical workunits, so we will only send out these jobs to computers that have >1Gb RAM.

(3) Also, we're trying a new feature where at the end of a simulation, Rosetta compares its fold to the predictions made by a dozen other algorithms. (Those predictions are sent to the clients in a compressed format.) Seeing consensus between different algorithms is usually a good sign that a prediction is right.

---