def doCovar(self, winsize=.1, slide=.05):
# Calculate Covariance matrices over all temporal data IAW given window param
self.covar = []
self.fealcov = []
for i, tr in self.E.trlist.items():
if i % 100 == 0:
print(i)
cov = dr.calc_covar(tr.xyz, winsize, 1, slide=slide)
if self.space in ['ds', 'dsw']:
X, C = dr.distance_space(tr), self.E.cent_ds
else:
X, C = tr.xyz, self.E.cent_c
wgt = self.E.cw if self.space in ['cw', 'dsw'] else [1,1,1,1,1]
rms = calc_rmsd(X, C, weights=wgt)
W = int(winsize * 1000)
S = int(slide * 1000)
feal = [np.mean([FL.feal.atemporal(i) for i in rms[st:st+W]], axis=0) for st in range(0, len(tr.xyz), S)]
for n in range(min(len(cov), len(feal))):
self.covar.append(cov[n])
self.fealcov.append(feal[n])
def calc_feal_AB(cls, traj):
"""Atemporal (individual frame) featue landscape
"""
maxd = self.max_rms_dist
mind = self.min_rms_dist
ds = DR.distance_space(traj)
rms = rmsd.calc_rmsd(ds, self.cent_ds)
# Proximity to State
for i in range(traj.n_frames):
fealand = [0 for i in range(5)]
fealand[np.argmin(rms[i])] = self.scaleto
# Proximity
for dist in enumerate(rms):
fealand.append(scaleto*max(maxd-(max(dist, mind), 0))/(maxd-mind))
# Additional Feature Spaces
for a in range(4):
for b in range(a+1, 5):
rel_dist = rms[a]-rms[b]
tup.append(log_reld(rel_dist))
fealand.extend(tup)
def calc_feal_AB(cls, traj):
"""Atemporal (individual frame) featue landscape
"""
maxd = self.max_rms_dist
mind = self.min_rms_dist
ds = DR.distance_space(traj)
rms = rmsd.calc_rmsd(ds, self.cent_ds)
# Proximity to State
for i in range(traj.n_frames):
fealand = [0 for i in range(5)]
fealand[np.argmin(rms[i])] = self.scaleto
# Proximity
for dist in enumerate(rms):
fealand.append(scaleto * max(maxd - (max(dist, mind), 0)) /
(maxd - mind))
# Additional Feature Spaces
for a in range(4):
for b in range(a + 1, 5):
rel_dist = rms[a] - rms[b]
tup.append(log_reld(rel_dist))
fealand.extend(tup)
traj, rmsd = [], []
for tnum in range(42):
p, d = DE.getHistoricalTrajectory_prot(tnum)
traj.append(md.load(d, top=p).atom_slice(afilt))
resrms = []
ccent = np.load('../data/bpti-alpha-cartesian-centroid.npy')
for tnum, tr in enumerate(detraj):
minlist = TS.TimeScape.read_log(home+'/work/timescape/desh_%02d_minima.log'%tnum)
minima = tr.slice(minlist)
minima.superpose(topoa)
for m in minima.xyz:
resrms.append(LA.norm(ref - m, axis=1))
ds = DR.distance_space(minima)
state = np.array([np.argmin(LA.norm(cent-i, axis=1)) for i in ds])
resrms = [np.zeros(58)]
for in range(len(minima[1:])):
resrms.append(LA.norm(minima[i].xyz - minima[i-1], axis=2))
for m, s in zip(minima, state):
resrms.append(LA.norm(m.xyz - ccent[s], axis=0))
for i in range(1, 91116):
CM[i] = np.abs(resrms[max(0, i-5):i].mean(0)-resrms[i: min(91116, i+5)].mean(0)) > theta
a, b, Z = 0, 30000, 300
for sc in range(8):
data = {'State':bL[a:b]}
for rd in range(7):
def bootstrap_lattice(catalog, num=10, build_new=False):
''' Bootstrap After TimeScape has run on source trajectory '''
home = os.getenv("HOME")
support = 1
cutoff = 8
start_coord = ['de2586_315', 'de531_20', 'de3765_63', 'de3305_668', 'de1732_139']
dcdfile = lambda x: home + '/work/data/{0}.dcd'.format(x)
outloc = lambda x: home+'/work/jc/denovouniform1/{0}/{0}'.format(x)
traj_list = {}
basin_list = catalog.lrange('basin:list', 0, -1)
if len(basin_list) == 134:
logging.info('Basin Data already loaded!')
rms_delta_list = [(i, np.sum(pickle.loads(catalog.get('basin:rmsdelta:'+b)))) for i, b in enumerate(basin_list)]
else:
logging.info('Loading all bootstrap data to initialize...')
basin_list = []
rms_delta_list = []
pdb_file = home+'/work/data/alpha.pdb'
topo = md.load(pdb_file)
ref_alpha = md.load(home+'/work/' + catalog.get('pdb:ref:0'))
ref_alpha.atom_slice(ref_alpha.top.select_atom_indices('alpha'), inplace=True)
res_rms_Kr = FEATURE_SET
for sc in start_coord:
dist_space = []
srcfile = outloc(sc) + '.dcd'
pdbfile = srcfile.replace('dcd', 'pdb')
logging.debug('LOADING TRAJ: %s', srcfile)
traj = md.load(srcfile, top = pdbfile)
traj_list[sc] = traj
alpha = traj.atom_slice(traj.top.select_atom_indices('alpha'))
logging.info('Grabbing TS data...')
W = TS.TimeScape.windows(outloc(sc) + '_transitions.log')
ts_traj = TS.TimeScapeParser(pdbfile, outloc(sc), sc, dcd=srcfile, traj=traj)
basins = ts_traj.load_basins()
logging.info("Processing distance space and residue RMS")
dsa = DR.distance_space(alpha)
resrmsd = 10*np.array([LA.norm(i-ref_alpha.xyz[0], axis=1) for i in alpha.xyz])
basin_res_rms = np.zeros(shape=(len(ts_traj.basins), alpha.n_atoms))
for i, (a,b) in enumerate(W):
dist_space.append(dsa[a:b].mean(0))
basin_res_rms[i] = np.median(resrmsd[a:b], axis=0)
basin_res_rms_delta = np.array([rms_delta(i) for i in basin_res_rms.T]).T
logging.debug('RMS LEN CHECK: %d =?= %d -- Updating RMS Delta',len(basins), len(basin_res_rms_delta))
for i, basin in enumerate(basins):
pipe = catalog.pipeline()
bid = basin.id
# Store on Disk and in redis
jc_filename = os.path.join(settings.datadir, 'basin_%s.pdb' % bid)
logging.info('MIN for %s: Idx# %d to %s', bid, basin.mindex, jc_filename)
minima_frame = traj.slice(basin.mindex) #md.load_frame(src_traj, basin.mindex, top=src_traj.replace('dcd', 'pdb'))
minima_frame.save_pdb(jc_filename)
basin_hash = basin.kv()
basin_hash['pdbfile'] = jc_filename
logging.info(' Basin: %(id)s %(start)d - %(end)d Minima: %(mindex)d size=%(len)d' % basin_hash)
pipe.rpush('basin:list', bid)
pipe.hmset('basin:%s'%bid, basin_hash)
pipe.set('basin:dmu:'+bid, pickle.dumps(dist_space[i]))
pipe.set('minima:%s'%bid, pickle.dumps(minima_frame))
# FOR RESIDUE RMSD
resrms_d = np.sum(basin_res_rms_delta[i][res_rms_Kr])
basin_hash['resrms_delta'] = resrms_d
rms_delta_list.append((len(basin_list), resrms_d))
basin_list.append(basin_hash)
pipe.set('basin:rmsdelta:'+bid, pickle.dumps(basin_res_rms_delta[i]))
pipe.execute()
# FOR RESIDUE RMSD
# FOR SEED SAMPLING USING RMS_DELTA
# Note: skip the first basin
# Re-Construct the Lattice from
if build_new:
dist_space = 10*np.array(dist_space)
cm = ds<cutoff
fs = lat.reduced_feature_set(cm,.115); len(fs)
dr, cr = ds[:,fs], cm[:,fs]
mfis,lfis = lat.maxminer(cr, 1)
dlat, ik = lat.derived_lattice(mfis, dr, cr)
pickle.dump(mfis, open(home + '/work/data/denovo_mfis.p', 'wb'))
pickle.dump(lfis, open(home + '/work/data/denovo_lfis.p', 'wb'))
pickle.dump(ik, open(home + '/work/data/denovo_iset.p', 'wb'))
pickle.dump(dlat, open(home + '/work/data/denovo_dlat.p', 'wb'))
else:
logging.info('Loading Pre-Constructed Lattice Data')
dlat = pickle.load(open(home + '/work/data/denovo_dlat.p', 'rb'))
mfis = pickle.load(open(home + '/work/data/denovo_mfis.p', 'rb'))
lfis = pickle.load(open(home + '/work/data/denovo_lfis.p', 'rb'))
ik = pickle.load(open(home + '/work/data/denovo_iset.p', 'rb'))
with catalog.pipeline() as pipe:
pipe.set('lattice:max_fis', pickle.dumps(mfis))
pipe.set('lattice:low_fis', pickle.dumps(lfis))
pipe.set('lattice:dlat', pickle.dumps(dlat))
pipe.set('lattice:iset', pickle.dumps(ik))
pipe.execute()
# logging.info('Building Existing lattice object')
# lattice=lat.Lattice(ds, fs, cutoff, support)
# lattice.set_fis(max_fis, low_fis)
# lattice.set_dlat(dlat, Ik)
# sampler = LatticeSampler(lattice)
# Sample -- FOR USING LATTICE TO BOOTSTRAP
# cl,sc,el = lat.clusterlattice(dlat, cr, dr, ik, num_k=8, invert=True)
# cl_list = sorted(el, key=lambda x: len(x))
# TODO: Check if fan out > single item clusters
# start_indices = [clu[0][0] for clu in cl_list[:num]]
rms_delta_ranked = [x[0] for x in sorted(rms_delta_list, key=lambda i: i[1], reverse=True)]
start_indices = rms_delta_ranked[:num]
seedlist = [catalog.lindex('basin:list', i) for i in start_indices]
sim_init = {key: catalog.get(key) for key in settings.sim_params.keys()}
global_params = getSimParameters(sim_init, 'seed')
global_params['psf'] = home+'/work/jc/serial2/de0_0/de0_0.psf'
for seed in seedlist:
logging.debug('\nSeeding Job: %s ', seed)
basin = catalog.hgetall('basin:%s'%seed)
catalog.rpush('executed', seed)
# Generate new set of params/coords
jcID, config = generateFromBasin(basin)
# Update Additional JC Params and Decision History, as needed
config.update(global_params)
# Push to catalog
logging.info("New Simulation Job Created: %s", jcID)
for k, v in config.items():
logging.debug(" %s: %s", k, str(v))
catalog.rpush('jcqueue', jcID)
catalog.hmset(wrapKey('jc', jcID), config)
def execute(self, job):
# PRE-PREOCESS ----------------------------------------------------------
settings = systemsettings()
bench = microbench("sim_%s" % settings.name, self.seqNumFromID())
bench.start()
stat = StatCollector("sim_%s" % settings.name, self.seqNumFromID())
mylogical_seqnum = str(self.seqNumFromID())
# Prepare working directory, input/output files
conFile = os.path.join(job["workdir"], job["name"] + ".conf")
logFile = conFile.replace("conf", "log") # log in same place as config file
dcdFile = conFile.replace("conf", "dcd") # dcd in same place as config file
USE_SHM = True
SIMULATE_RATIO = settings.SIMULATE_RATIO
if SIMULATE_RATIO > 1:
logging.warning(" USING SIMULATION RATIO OF %d -- THis is ONLY for debugging", SIMULATE_RATIO)
frame_size = (SIMULATE_RATIO * int(job["interval"])) / (1000)
logging.info("Frame Size is %f Using Sim Ratio of 1:%d", frame_size, SIMULATE_RATIO)
EXPERIMENT_NUMBER = settings.EXPERIMENT_NUMBER
logging.info("Running Experiment Configuration #%d", EXPERIMENT_NUMBER)
# # Grab historical basin data's relative time to start (for lineage)
traj = None
# EXECUTE SIMULATION ---------------------------------------------------------
if self.skip_simulation:
logging.info("1. SKIPPING SIMULATION.....")
USE_SHM = False
job["dcd"] = dcdFile
key = wrapKey("jc", job["name"])
self.data[key]["dcd"] = dcdFile
else:
logging.info("1. Run Simulation")
# Prepare & source to config file
with open(self.data["sim_conf_template"], "r") as template:
source = template.read()
# >>>>Storing DCD into shared memory on this node
if USE_SHM:
ramdisk = gettempdir()
job["outputloc"] = ramdisk
dcd_ramfile = os.path.join(ramdisk, job["name"] + ".dcd")
else:
job["outputloc"] = ""
with open(conFile, "w") as sysconfig:
sysconfig.write(source % job)
logging.info("Config written to: " + conFile)
# # Run simulation in parallel
# if 'parallel' in job:
# numnodes = job['parallel']
# total_tasks = numnodes * 24
# cmd = 'mpiexec -n %d namd2 %s > %s' % (total_tasks, conFile, logFile)
# # Run simulation single threaded
# else:
# cmd = 'namd2 %s > %s' % (conFile, logFile)
# cmd = 'mpirun -n %d namd2 %s > %s' % (PARALLELISM, conFile, logFile)
check = executecmd("module list")
logging.debug("%s", check)
cmd = "namd2 +p%d %s > %s" % (PARALLELISM, conFile, logFile)
# MICROBENCH #1 (file to Lustre)
# logging.debug("Executing Simulation:\n %s\n", cmd)
# bench = microbench()
# bench.start()
# stdout = executecmd(cmd)
# logging.info("SIMULATION Complete! STDOUT/ERR Follows:")
# bench.mark('SimExec:%s' % job['name'])
# shm_contents = os.listdir('/dev/shm/out')
# logging.debug('Ramdisk contents (should have files) : %s', str(shm_contents))
# shutil.copy(ramdisk + job['name'] + '.dcd', job['workdir'])
# logging.info("Copy Complete to Lustre.")
# bench.mark('CopyLustre:%s' % job['name'])
# shutil.rmtree(ramdisk)
# shm_contents = os.listdir('/dev/shm')
# logging.debug('Ramdisk contents (should be empty) : %s', str(shm_contents))
# bench.show()
max_expected_obs = int(job["runtime"]) // int(job["dcdfreq"])
# Retry upto 3 attempts if the sim fails
MAX_TRY = 3
for i in range(MAX_TRY, 0, -1):
min_required_obs = int(max_expected_obs * ((i - 1) / (MAX_TRY)))
logging.debug("Executing Simulation:\n %s\n", cmd)
logging.debug("# Obs Expected to see: %d", max_expected_obs)
stdout = executecmd(cmd)
logging.info("SIMULATION Complete! STDOUT/ERR Follows:")
# Check file for expected data
if USE_SHM:
traj = md.load(dcd_ramfile, top=job["pdb"])
else:
traj = md.load(dcdFile, top=job["pdb"])
logging.info("Obs Threshold = %4d", min_required_obs)
logging.info("#Obs This Traj = %4d", traj.n_frames)
if traj.n_frames >= min_required_obs:
logging.info("Full (enough) Sim Completed")
break
logging.info("Detected a failed Simulation. Retrying the same sim.")
break
bench.mark("simulation")
# bench.mark('SimExec:%s' % job['name'])
# Internal stats
sim_length = self.data["sim_step_size"] * int(job["runtime"])
sim_realtime = bench.delta_last()
sim_run_ratio = (sim_realtime / 60) / (sim_length / 1000000)
logging.info("##SIM_RATIO %6.3f min-per-ns-sim", sim_run_ratio)
stat.collect("sim_ratio", sim_run_ratio)
if USE_SHM:
shm_contents = os.listdir(ramdisk)
logging.debug("Ramdisk contents (should have files) : %s", str(shm_contents))
if not os.path.exists(dcd_ramfile):
logging.warning("DCD FILE NOT FOUND!!!! Wait 10 seconds for sim to close it (???)")
time.sleep(10)
if not os.path.exists(dcd_ramfile):
logging.warning("DCD STIILL FILE NOT FOUND!!!!")
else:
logging.info("DCD File was found")
# # MICROBENCH #2 (file to Alluxio)
# allux = AlluxioClient()
# # copy to Aluxio FS
# allux.put(ramdisk + job['name'] + '.dcd', '/')
# logging.info("Copy Complete to Alluxio.")
# bench.mark('CopyAllux:%s' % job['name'])
# And copy to Lustre
# shutil.copy(ramdisk + job['name'] + '.dcd', job['workdir'])
# And copy to Lustre (usng zero-copy):
if USE_SHM:
# ALT: X-Mit to Cache Service and let cache write to disk lazily
src = open(dcd_ramfile, "rb")
dest = open(dcdFile, "w+b")
offset = 0
dcdfilesize = os.path.getsize(dcd_ramfile)
while True:
sent = sendfile(dest.fileno(), src.fileno(), offset, dcdfilesize)
if sent == 0:
break
offset += sent
logging.info("Copy Complete to Lustre.")
bench.mark("CopyLustre:%s" % job["name"])
# TODO: Update job's metadata
key = wrapKey("jc", job["name"])
self.data[key]["dcd"] = dcdFile
# ANALYSIS ------- ---------------------------------------------------------
# ANALYSIS ALGORITHM
# 1. With combined Sim-analysis: file is loaded locally from shared mem
logging.debug("2. Load DCD")
# Load topology and define filters
topo = self.protein.pdb
pdb = md.load(job["pdb"])
hfilt = pdb.top.select_atom_indices("heavy")
pfilt = pdb.top.select("protein")
afilt = pdb.top.select_atom_indices("alpha")
# Load full higher dim trajectory
if traj is None:
if USE_SHM:
traj = md.load(dcd_ramfile, top=pdb)
else:
traj = md.load(dcdFile, top=pdb)
logging.debug("Trajectory Loaded: %s (%s)", job["name"], str(traj))
traj_prot = traj.atom_slice(pfilt)
traj_heavy = traj.atom_slice(hfilt)
traj_alpha = traj.atom_slice(afilt)
# Superpose Coordinates to Common Reference
traj_prot.superpose(topo)
# Calculate output Distance Space
# Use of the second side chain atom is discussed in the ref paper on Timescape
# The atom pairs and selected atoms are in the timescape module
sc_pairs = side_chain_pairs(traj_prot)
n_features = len(sc_pairs)
# Use the CA atoms to calculate distance space
# NOTE THE CONVERSION FROM NM TO ANG!!!!!!
dist_space = 10 * DR.distance_space(traj_alpha)
# Get the frame rate to for conversion between source and timescape
# NOTE: Timescae is run at a more coarse level
logging.debug("Preprocessing output for TimeScapes: terrain")
traj_frame_per_ps = SIMULATE_RATIO * int(job["interval"]) / 1000.0 # jc interval is in fs
ts_frame_per_ps = int(self.data["timescape:rate"]) # this value is in ps
frame_rate = int(ts_frame_per_ps / traj_frame_per_ps)
logging.debug(
"%5.2f fr/ps (Traj) %5.2f fr/ps (TS) FrameRate= %4.1f",
traj_frame_per_ps,
ts_frame_per_ps,
frame_rate,
)
# Execute Timescapes agility program to detect spatial-temporal basins
output_prefix = os.path.join(job["workdir"], job["name"])
if not self.skip_timescape:
# Prep file and save locally in shm
tmploc = gettempdir()
ts_out = tmploc + "traj_ts"
ts_dcd = ts_out + ".dcd"
ts_pdb = ts_out + ".pdb"
heavy = traj_heavy.slice(range(0, traj.n_frames, frame_rate))
heavy.slice(0).save_pdb(ts_pdb)
heavy.save_dcd(ts_dcd)
# Gaussuan Full Width at Half-Max value affects sliding window size
# ref: http://timescapes.biomachina.org/guide.pdf
gmd_cut1 = int(self.data["timescape:gmd:low"])
gmd_cut2 = int(self.data["timescape:gmd:hi"])
gauss_wght_delta = int(self.data["timescape:delta"])
# Execute timescapes' terrain.py on the pre-processed trajectory
cmd = "terrain.py %s %s %d %d %d GMD %s" % (
ts_pdb,
ts_dcd,
gmd_cut1,
gmd_cut2,
gauss_wght_delta,
output_prefix,
)
logging.info("Running Timescapes:\n %s", cmd)
stdout = executecmd(cmd)
logging.info("TimeScapes COMPLETE:\n%s", stdout)
# Collect and parse Timescape output
logging.debug("Parsing Timescapes output")
ts_parse = TimeScapeParser(job["pdb"], output_prefix, job["name"], dcd=dcdFile, traj=traj, uniqueid=False)
basin_list = ts_parse.load_basins(frame_ratio=frame_rate)
n_basins = len(basin_list)
minima_coords = {}
basin_rms = {}
basins = {}
new_dmu = {}
new_dsig = {}
resid_rms_delta = {}
stat.collect("num_basin", n_basins)
downstream_list = []
# FOR BOOTSTRAPPING USING RMSD
ref_file = os.path.join(settings.workdir, self.data["pdb:ref:0"])
logging.info("Loading RMSD reference frame from %s", self.data["pdb:ref:0"])
refframe = md.load(ref_file)
ref_alpha = refframe.atom_slice(refframe.top.select_atom_indices("alpha"))
rmsd = 10 * md.rmsd(traj_alpha, ref_alpha)
# FOR RESIDUE RMSD
res_rms_Kr = FEATURE_SET_RESID
resrmsd = 10 * np.array([LA.norm(i - ref_alpha.xyz[0], axis=1) for i in traj_alpha.xyz])
basin_res_rms = np.zeros(shape=(len(basin_list), traj_alpha.n_atoms))
# LOAD CENROIDS -- todo move to immut
centroid = pickle.loads(self.catalog.get("centroid:ds"))
basin_label_list = {}
# Process each basin
for i, basin in enumerate(basin_list):
logging.info(" Processing basin #%2d", i)
bid = basin.id
downstream_list.append(bid)
# Slice out minima coord & save to disk (for now)
# TODO: Store in memory in cache
minima_coords[bid] = traj.slice(basin.mindex)
jc_filename = os.path.join(settings.datadir, "basin_%s.pdb" % bid)
minima_coords[bid].save_pdb(jc_filename)
# METRIC CALCULATION
a, b = basin.start, basin.end
new_dmu[bid] = np.mean(dist_space[a:b], axis=0)
new_dsig[bid] = np.std(dist_space[a:b], axis=0)
# Collect Basin metadata
basin_hash = basin.kv()
basin_hash["pdbfile"] = jc_filename
basin_hash["rmsd"] = np.median(rmsd[a:b])
basin_res_rms[i] = np.median(resrmsd[a:b], axis=0)
# FOR EXP #16 -- Residue RMSD
# Set relative time (in ns)
# basin_hash['time'] = reltime_start + (a * frame_rate) / 1000
basins[bid] = basin_hash
# LABEL ATEMPORAL DATA
# Take every 4th frame
stride = [dist_space[i] for i in range(a, b, 4)]
rms = [LA.norm(centroid - i, axis=1) for i in stride]
label_seq = [np.argmin(i) for i in rms]
basin_label_list[bid] = label_seq
logging.info(" Basin Processed: #%s, %d - %d", basin_hash["traj"], basin_hash["start"], basin_hash["end"])
# RMSD DELTA (RESIDUE)
basin_res_rms_delta = np.array([rms_delta(i) for i in basin_res_rms.T]).T
basin_rms_delta_bykey = {basin.id: basin_res_rms_delta[i] for i, basin in enumerate(basin_list)}
for k in basins.keys():
basins[k]["resrms_delta"] = np.sum(basin_rms_delta_bykey[k][res_rms_Kr])
# TODO: Use Min Index as snapshot, median (or mean) DistSpace vals for each basin?????
bench.mark("analysis")
# BARRIER: WRITE TO CATALOG HERE -- Ensure Catalog is available
# try:
self.wait_catalog()
# except OverlayNotAvailable as e:
# logging.warning("Catalog Overlay Service is not available. Scheduling ASYNC Analysis")
# FOR EXPERIMENT METRICS
src_basin = self.catalog.hgetall("basin:" + job["src_basin"])
with open("/home-1/[email protected]/ddc/results/{0}_prov.log".format(settings.name), "a") as metric_out:
for i, basin in enumerate(basin_list):
bid = basin.id
label_seq = basin_label_list[bid]
basin_metric_label = LABEL10(label_seq)
metric_out.write(
"BASIN,%s,%s,%s,%s\n"
% (bid, src_basin["label:10"], basin_metric_label, "".join([str(i) for i in label_seq]))
)
# Update Index Synchronized data lists
basin_list_sorted = sorted(basins.keys(), key=lambda x: (x.split("_")[0], int(x.split("_")[1])))
for bid in basin_list_sorted:
with self.catalog.pipeline() as pipe:
while True:
try:
logging.debug("Updating %s basin indeces and distance space", len(basins))
pipe.rpush("basin:list", bid)
pipe.rpush("dspace", pickle.dumps(new_dmu[bid]))
basin_index, _ = pipe.execute()
break
except redis.WatchError as e:
logging.debug("WATCH ERROR. Someone else is writing to the catalog. Retrying...")
continue
basins[bid]["dsidx"] = basin_index - 1
# Update Catalog with 1 Long Atomic Transaction
with self.catalog.pipeline() as pipe:
while True:
try:
logging.debug("Update Filelist")
pipe.watch(wrapKey("jc", job["name"]))
file_idx = pipe.rpush("xid:filelist", job["dcd"]) - 1
# HD Points
logging.debug("Update HD Points")
start_index = pipe.llen("xid:reference")
pipe.multi()
pipe.rpush("xid:reference", *[(file_idx, x) for x in range(traj.n_frames)])
pipe.set("resrms:" + job["name"], resrmsd)
# Store all basin data
logging.debug("Updating %s basins", len(basins))
for bid in sorted(basins.keys(), key=lambda x: (x.split("_")[0], int(x.split("_")[1]))):
pipe.hmset("basin:" + bid, basins[bid])
pipe.set("minima:%s" % bid, pickle.dumps(minima_coords[bid]))
for i in basin_label_list[bid]:
pipe.rpush("basin:labelseq:" + bid, i)
pipe.hset("anl_sequence", job["name"], mylogical_seqnum)
logging.debug("Executing")
pipe.execute()
break
except redis.WatchError as e:
logging.debug("WATCH ERROR. Someone else is writing to the catalog. Retrying...")
continue
self.data[key]["xid:start"] = start_index
self.data[key]["xid:end"] = start_index + traj.n_frames
bench.mark("catalog")
# ---- POST PROCESSING
# Check for total data in downstream queue & Notify control manager to run
# 1. Det. what's still running
job_queue = slurm.currentqueue()
n_simjobs = -1 # -1 to exclude self
for j in job_queue:
if j["name"].startswith("sw"):
n_simjobs += 1
logging.info("SIM WORKER has detected %d other simulations running/pending.", len(job_queue))
remain_percent = n_simjobs / self.data["numresources"]
# Fault Tolerance: ensure the pipeline persists
if n_simjobs == 0:
logging.info("I am the last simulation. Ensuring the controler executes.")
self.catalog.set("ctl:force", 1)
# Send downstream notification when less than 10% of jobs are still running
if not self.skip_notify and remain_percent < 0.1:
# Notify the control manager 'CM'
self.notify("ctl")
if USE_SHM:
shutil.rmtree(ramdisk)
shm_contents = os.listdir("/dev/shm")
logging.debug("Ramdisk contents (should be empty of DDC) : %s", str(shm_contents))
# For benchmarching:
bench.show()
stat.show()
# Return # of observations (frames) processed
return downstream_list
all_label = np.zeros(shape=(len(jclist), 4500))
sumlog = open(home + '/work/results/trajout_{0}'.format(expname), 'w')
for i, jc in enumerate(jclist[startnum:]):
if 'dcd' not in jc:
jc['dcd'] = jc['pdb'].replace('dcd', 'pdb')
if not os.path.exists(jc['dcd']):
logging.info("%d %s NO_DCD", i, jc['name'])
continue
tr = md.load(jc['dcd'], top=topo, stride=4)
if tr.n_frames < 4503:
logging.info("%d %s Frames: %d", i, jc['name'], tr.n_frames)
continue
prot = tr.atom_slice(pfilt)
prot.save_dcd(wkdir + '/%s_%03d.dcd' % (expname, (i + startnum)))
alpha = prot.atom_slice(afilt)
ds = DR.distance_space(alpha)
rms = [LA.norm(cent-i, axis=1) for i in ds]
label = np.array([np.argmin(i) for i in rms[3:]])
all_label[i] = label
label_str = ''.join([str(i) for i in label])
red.hset('basin:' + jc['name'], 'label', label_str)
bincnt = ','.join([str(i) for i in np.bincount(label, minlength=5)])
src_basin = jc['src_basin'] if 'src_basin' in jc else 'NONE'
logging.info("%d,%s,%s,%s", i, jc['name'], src_basin, bincnt)
sumlog.write('%d,%s,%s,%s\n' % (i, jc['name'], src_basin, bincnt))
sumlog.close()
np.save(home + '/work/results/label_{0}'.format(expname), all_label)
logging.info('ALL Done!')
def load_historical_Expl(catalog):
""" Load all DEShaw data into basins for processing """
settings = systemsettings()
# idx_list = [0,1,2,3,4,20,23,24,30,32,34,40,41,42]
idx_list = [0,34]
tlist = {k: 'tr%d.dcd'%k for k in idx_list}
seed_dir = os.path.join(settings.WORKDIR, 'seed')
seed_ts_ratio = 16 # TimeScape ran on 4ps frame rate (16x source)
# Load topology and anscillary data
# bpti = Protein(bpti, catalog, load=True)
pdb_file = os.path.join(seed_dir, 'coord.pdb')
topo = md.load(pdb_file)
pfilt = topo.top.select('protein')
logging.info('Topology loaded %s', topo)
# ID Side Chain pair atoms for each distance space calc
sc_pairs = side_chain_pairs(topo.atom_slice(pfilt))
logging.info('Identified side chains: %d', len(sc_pairs))
# Process all sorce trajectories
basin_list = []
C_T, mu_T, sigma_T = [], [], []
for idx in idx_list:
logging.info('Procesing Seed index: %d', idx)
# Load SRC seed trajetory & calc distance space -- TODO: make this optional
tfile = os.path.join(seed_dir, tlist[idx])
traj = md.load(tfile, top=topo)
traj.superpose(topo)
ds = datareduce.distance_space(traj, pairs=sc_pairs)
# Push to Catalog
file_idx = catalog.rpush('xid:filelist', tfile) - 1
start_index = catalog.llen('xid:reference')
# TODO: Do I still need to index every frame??????
catalog.rpush('xid:reference', *[(file_idx, x) for x in range(traj.n_frames)])
# Process Trajectory as basins
logging.info(" Seed Loaded. Loading TimeScape Data...")
seed_name = 'tr%d'%idx
ts_data_path = os.path.join(seed_dir, 'TEST', seed_name)
ts_traj = TimeScapeParser(pdb_file,
ts_data_path, seed_name, dcd=tfile, traj=traj)
basin_list = ts_traj.load_basins(frame_ratio=seed_ts_ratio)
corr_mat = ts_traj.correlation_matrix()
for i, basin in enumerate(ts_traj.basins):
a, b = basin.start, basin.end
bid = basin.id
if a > traj.n_frames:
logging.info('Finished processing all basins for this Trajectory!')
break
# Store on Disk and in redis
jc_filename = os.path.join(settings.datadir, 'basin_%s.pdb' % bid)
minima_frame = md.load_frame(tfile, basin.mindex, top=topo) if traj is None else traj.slice(basin.mindex)
minima_frame.save_pdb(jc_filename)
C_T.append(np.mean(corr_mat[a:b], axis=0))
mu_T.append(np.mean(ds[a:b], axis=0))
sigma_T.append(np.std(ds[a:b], axis=0))
basin_hash = basin.kv()
basin_hash['pdbfile'] = jc_filename
basin_hash['dsidx'] = i
logging.info(' Basin: %(id)s %(start)d - %(end)d Minima: %(mindex)d size=%(len)d' % basin_hash)
pipe = catalog.pipeline()
pipe.rpush('basin:list', bid)
pipe.hmset('basin:%s'%bid, basin_hash)
pipe.set('minima:%s'%bid, pickle.dumps(minima_frame))
pipe.execute()
catalog.set('basin:processed', len(ts_traj.basins))
catalog.storeNPArray(np.array(C_T), 'corr_vector')
catalog.storeNPArray(np.array(mu_T), 'dspace_mu')
def execute(self, job):
# PRE-PREOCESS ----------------------------------------------------------
settings = systemsettings()
bench = microbench('sim_%s' % settings.name, self.seqNumFromID())
bench.start()
stat = StatCollector('sim_%s' % settings.name, self.seqNumFromID())
mylogical_seqnum = str(self.seqNumFromID())
# Prepare working directory, input/output files
conFile = os.path.join(job['workdir'], job['name'] + '.conf')
logFile = conFile.replace('conf', 'log') # log in same place as config file
dcdFile = conFile.replace('conf', 'dcd') # dcd in same place as config file
USE_SHM = True
SIMULATE_RATIO = settings.SIMULATE_RATIO
if SIMULATE_RATIO > 1:
logging.warning(" USING SIMULATION RATIO OF %d -- THis is ONLY for debugging", SIMULATE_RATIO)
frame_size = (SIMULATE_RATIO * int(job['interval'])) / (1000)
logging.info('Frame Size is %f Using Sim Ratio of 1:%d', \
frame_size, SIMULATE_RATIO)
EXPERIMENT_NUMBER = settings.EXPERIMENT_NUMBER
logging.info('Running Experiment Configuration #%d', EXPERIMENT_NUMBER)
# # Grab historical basin data's relative time to start (for lineage)
traj = None
# EXECUTE SIMULATION ---------------------------------------------------------
if self.skip_simulation:
logging.info('1. SKIPPING SIMULATION.....')
USE_SHM = False
job['dcd'] = dcdFile
key = wrapKey('jc', job['name'])
self.data[key]['dcd'] = dcdFile
else:
logging.info('1. Run Simulation')
# Prepare & source to config file
with open(self.data['sim_conf_template'], 'r') as template:
source = template.read()
# >>>>Storing DCD into shared memory on this node
if USE_SHM:
ramdisk = gettempdir()
job['outputloc'] = ramdisk
dcd_ramfile = os.path.join(ramdisk, job['name'] + '.dcd')
else:
job['outputloc'] = ''
with open(conFile, 'w') as sysconfig:
sysconfig.write(source % job)
logging.info("Config written to: " + conFile)
# # Run simulation in parallel
# if 'parallel' in job:
# numnodes = job['parallel']
# total_tasks = numnodes * 24
# cmd = 'mpiexec -n %d namd2 %s > %s' % (total_tasks, conFile, logFile)
# # Run simulation single threaded
# else:
# cmd = 'namd2 %s > %s' % (conFile, logFile)
# cmd = 'mpirun -n %d namd2 %s > %s' % (PARALLELISM, conFile, logFile)
check = executecmd('module list')
logging.debug('%s', check)
cmd = 'namd2 +p%d %s > %s' % (PARALLELISM, conFile, logFile)
# MICROBENCH #1 (file to Lustre)
# logging.debug("Executing Simulation:\n %s\n", cmd)
# bench = microbench()
# bench.start()
# stdout = executecmd(cmd)
# logging.info("SIMULATION Complete! STDOUT/ERR Follows:")
# bench.mark('SimExec:%s' % job['name'])
# shm_contents = os.listdir('/dev/shm/out')
# logging.debug('Ramdisk contents (should have files) : %s', str(shm_contents))
# shutil.copy(ramdisk + job['name'] + '.dcd', job['workdir'])
# logging.info("Copy Complete to Lustre.")
# bench.mark('CopyLustre:%s' % job['name'])
# shutil.rmtree(ramdisk)
# shm_contents = os.listdir('/dev/shm')
# logging.debug('Ramdisk contents (should be empty) : %s', str(shm_contents))
# bench.show()
max_expected_obs = int(job['runtime']) // int(job['dcdfreq'])
# Retry upto 3 attempts if the sim fails
MAX_TRY = 3
for i in range(MAX_TRY, 0, -1):
min_required_obs = int(max_expected_obs * ((i-1)/(MAX_TRY)))
logging.debug("Executing Simulation:\n %s\n", cmd)
logging.debug('# Obs Expected to see: %d', max_expected_obs)
stdout = executecmd(cmd)
logging.info("SIMULATION Complete! STDOUT/ERR Follows:")
# Check file for expected data
if USE_SHM:
traj = md.load(dcd_ramfile, top=job['pdb'])
else:
traj = md.load(dcdFile, top=job['pdb'])
logging.info("Obs Threshold = %4d", min_required_obs)
logging.info("#Obs This Traj = %4d", traj.n_frames)
if traj.n_frames >= min_required_obs:
logging.info('Full (enough) Sim Completed')
break
logging.info('Detected a failed Simulation. Retrying the same sim.')
break
bench.mark('simulation')
# bench.mark('SimExec:%s' % job['name'])
# Internal stats
sim_length = self.data['sim_step_size'] * int(job['runtime'])
sim_realtime = bench.delta_last()
sim_run_ratio = (sim_realtime/60) / (sim_length/1000000)
logging.info('##SIM_RATIO %6.3f min-per-ns-sim', sim_run_ratio)
stat.collect('sim_ratio', sim_run_ratio)
if USE_SHM:
shm_contents = os.listdir(ramdisk)
logging.debug('Ramdisk contents (should have files) : %s', str(shm_contents))
if not os.path.exists(dcd_ramfile):
logging.warning("DCD FILE NOT FOUND!!!! Wait 10 seconds for sim to close it (???)")
time.sleep(10)
if not os.path.exists(dcd_ramfile):
logging.warning("DCD STIILL FILE NOT FOUND!!!!")
else:
logging.info("DCD File was found")
# # MICROBENCH #2 (file to Alluxio)
# allux = AlluxioClient()
# # copy to Aluxio FS
# allux.put(ramdisk + job['name'] + '.dcd', '/')
# logging.info("Copy Complete to Alluxio.")
# bench.mark('CopyAllux:%s' % job['name'])
# And copy to Lustre
# shutil.copy(ramdisk + job['name'] + '.dcd', job['workdir'])
# And copy to Lustre (usng zero-copy):
if USE_SHM:
# ALT: X-Mit to Cache Service and let cache write to disk lazily
src = open(dcd_ramfile, 'rb')
dest = open(dcdFile, 'w+b')
offset = 0
dcdfilesize = os.path.getsize(dcd_ramfile)
while True:
sent = sendfile(dest.fileno(), src.fileno(), offset, dcdfilesize)
if sent == 0:
break
offset += sent
logging.info("Copy Complete to Lustre.")
bench.mark('CopyLustre:%s' % job['name'])
# TODO: Update job's metadata
key = wrapKey('jc', job['name'])
self.data[key]['dcd'] = dcdFile
# ANALYSIS ------- ---------------------------------------------------------
# ANALYSIS ALGORITHM
# 1. With combined Sim-analysis: file is loaded locally from shared mem
logging.debug("2. Load DCD")
# Load topology and define filters
topo = self.protein.pdb
pdb = md.load(job['pdb'])
hfilt = pdb.top.select_atom_indices('heavy')
pfilt = pdb.top.select('protein')
afilt = pdb.top.select_atom_indices('alpha')
# Load full higher dim trajectory
if traj is None:
if USE_SHM:
traj = md.load(dcd_ramfile, top=pdb)
else:
traj = md.load(dcdFile, top=pdb)
logging.debug('Trajectory Loaded: %s (%s)', job['name'], str(traj))
traj_prot = traj.atom_slice(pfilt)
traj_heavy = traj.atom_slice(hfilt)
traj_alpha = traj.atom_slice(afilt)
# Superpose Coordinates to Common Reference
traj_prot.superpose(topo)
# Calculate output Distance Space
# Use of the second side chain atom is discussed in the ref paper on Timescape
# The atom pairs and selected atoms are in the timescape module
sc_pairs = side_chain_pairs(traj_prot)
n_features = len(sc_pairs)
# Use the CA atoms to calculate distance space
# NOTE THE CONVERSION FROM NM TO ANG!!!!!!
dist_space = 10 * DR.distance_space(traj_alpha)
# Get the frame rate to for conversion between source and timescape
# NOTE: Timescae is run at a more coarse level
logging.debug('Preprocessing output for TimeScapes: terrain')
traj_frame_per_ps = SIMULATE_RATIO * int(job['interval']) / 1000. # jc interval is in fs
ts_frame_per_ps = int(self.data['timescape:rate']) # this value is in ps
frame_rate = int(ts_frame_per_ps / traj_frame_per_ps)
logging.debug('%5.2f fr/ps (Traj) %5.2f fr/ps (TS) FrameRate= %4.1f', traj_frame_per_ps, ts_frame_per_ps, frame_rate)
# Execute Timescapes agility program to detect spatial-temporal basins
output_prefix = os.path.join(job['workdir'], job['name'])
if not self.skip_timescape:
# Prep file and save locally in shm
tmploc = gettempdir()
ts_out = tmploc + 'traj_ts'
ts_dcd = ts_out + '.dcd'
ts_pdb = ts_out + '.pdb'
heavy = traj_heavy.slice(range(0, traj.n_frames, frame_rate))
heavy.slice(0).save_pdb(ts_pdb)
heavy.save_dcd(ts_dcd)
# Gaussuan Full Width at Half-Max value affects sliding window size
# ref: http://timescapes.biomachina.org/guide.pdf
gmd_cut1 = int(self.data['timescape:gmd:low'])
gmd_cut2 = int(self.data['timescape:gmd:hi'])
gauss_wght_delta = int(self.data['timescape:delta'])
# Execute timescapes' terrain.py on the pre-processed trajectory
cmd = 'terrain.py %s %s %d %d %d GMD %s' %\
(ts_pdb, ts_dcd, gmd_cut1, gmd_cut2, gauss_wght_delta, output_prefix)
logging.info('Running Timescapes:\n %s', cmd)
stdout = executecmd(cmd)
logging.info('TimeScapes COMPLETE:\n%s', stdout)
# Collect and parse Timescape output
logging.debug('Parsing Timescapes output')
ts_parse = TimeScapeParser(job['pdb'], output_prefix, job['name'],
dcd=dcdFile, traj=traj, uniqueid=False)
basin_list = ts_parse.load_basins(frame_ratio=frame_rate)
n_basins = len(basin_list)
minima_coords = {}
basin_rms = {}
basins = {}
new_dmu={}
new_dsig={}
resid_rms_delta = {}
stat.collect('num_basin', n_basins)
downstream_list = []
# FOR BOOTSTRAPPING USING RMSD
ref_file = os.path.join(settings.workdir, self.data['pdb:ref:0'])
logging.info('Loading RMSD reference frame from %s', self.data['pdb:ref:0'])
refframe = md.load(ref_file)
ref_alpha = refframe.atom_slice(refframe.top.select_atom_indices('alpha'))
rmsd = 10*md.rmsd(traj_alpha, ref_alpha)
# FOR RESIDUE RMSD
res_rms_Kr = FEATURE_SET_RESID
resrmsd = 10*np.array([LA.norm(i-ref_alpha.xyz[0], axis=1) for i in traj_alpha.xyz])
basin_res_rms = np.zeros(shape=(len(basin_list), traj_alpha.n_atoms))
# LOAD CENROIDS -- todo move to immut
centroid = pickle.loads(self.catalog.get('centroid:ds'))
basin_label_list = {}
# Process each basin
for i, basin in enumerate(basin_list):
logging.info(' Processing basin #%2d', i)
bid = basin.id
downstream_list.append(bid)
# Slice out minima coord & save to disk (for now)
# TODO: Store in memory in cache
minima_coords[bid] = traj.slice(basin.mindex)
jc_filename = os.path.join(settings.datadir, 'basin_%s.pdb' % bid)
minima_coords[bid].save_pdb(jc_filename)
# METRIC CALCULATION
a, b = basin.start, basin.end
new_dmu[bid] = np.mean(dist_space[a:b], axis=0)
new_dsig[bid] = np.std(dist_space[a:b], axis=0)
# Collect Basin metadata
basin_hash = basin.kv()
basin_hash['pdbfile'] = jc_filename
basin_hash['rmsd'] = np.median(rmsd[a:b])
basin_res_rms[i] = np.median(resrmsd[a:b], axis=0)
# FOR EXP #16 -- Residue RMSD
# Set relative time (in ns)
# basin_hash['time'] = reltime_start + (a * frame_rate) / 1000
basins[bid] = basin_hash
# LABEL ATEMPORAL DATA
# Take every 4th frame
stride = [dist_space[i] for i in range(a,b,4)]
rms = [LA.norm(centroid - i, axis=1) for i in stride]
label_seq = [np.argmin(i) for i in rms]
basin_label_list[bid] = label_seq
logging.info(' Basin Processed: #%s, %d - %d', basin_hash['traj'],
basin_hash['start'], basin_hash['end'])
# RMSD DELTA (RESIDUE)
basin_res_rms_delta = np.array([rms_delta(i) for i in basin_res_rms.T]).T
basin_rms_delta_bykey = {basin.id: basin_res_rms_delta[i] for i, basin in enumerate(basin_list)}
for k in basins.keys():
basins[k]['resrms_delta'] = np.sum(basin_rms_delta_bykey[k][res_rms_Kr])
# TODO: Use Min Index as snapshot, median (or mean) DistSpace vals for each basin?????
bench.mark('analysis')
# BARRIER: WRITE TO CATALOG HERE -- Ensure Catalog is available
# try:
self.wait_catalog()
# except OverlayNotAvailable as e:
# logging.warning("Catalog Overlay Service is not available. Scheduling ASYNC Analysis")
# FOR EXPERIMENT METRICS
src_basin = self.catalog.hgetall("basin:" + job['src_basin'])
with open('/home-1/[email protected]/ddc/results/{0}_prov.log'.format(settings.name), 'a') as metric_out:
for i, basin in enumerate(basin_list):
bid = basin.id
label_seq = basin_label_list[bid]
basin_metric_label = LABEL10(label_seq)
metric_out.write('BASIN,%s,%s,%s,%s\n'% \
(bid, src_basin['label:10'], basin_metric_label, ''.join([str(i) for i in label_seq])))
# Update Index Synchronized data lists
basin_list_sorted =sorted(basins.keys(), key=lambda x: (x.split('_')[0], int(x.split('_')[1])))
for bid in basin_list_sorted:
with self.catalog.pipeline() as pipe:
while True:
try:
logging.debug('Updating %s basin indeces and distance space', len(basins))
pipe.rpush('basin:list', bid)
pipe.rpush('dspace', pickle.dumps(new_dmu[bid]))
basin_index,_ = pipe.execute()
break
except redis.WatchError as e:
logging.debug('WATCH ERROR. Someone else is writing to the catalog. Retrying...')
continue
basins[bid]['dsidx'] = basin_index - 1
# Update Catalog with 1 Long Atomic Transaction
with self.catalog.pipeline() as pipe:
while True:
try:
logging.debug('Update Filelist')
pipe.watch(wrapKey('jc', job['name']))
file_idx = pipe.rpush('xid:filelist', job['dcd']) - 1
# HD Points
logging.debug('Update HD Points')
start_index = pipe.llen('xid:reference')
pipe.multi()
pipe.rpush('xid:reference', *[(file_idx, x) for x in range(traj.n_frames)])
pipe.set('resrms:' + job['name'], resrmsd)
# Store all basin data
logging.debug('Updating %s basins', len(basins))
for bid in sorted(basins.keys(), key=lambda x: (x.split('_')[0], int(x.split('_')[1]))):
pipe.hmset('basin:'+bid, basins[bid])
pipe.set('minima:%s'%bid, pickle.dumps(minima_coords[bid]))
for i in basin_label_list[bid]:
pipe.rpush('basin:labelseq:'+bid, i)
pipe.hset('anl_sequence', job['name'], mylogical_seqnum)
logging.debug('Executing')
pipe.execute()
break
except redis.WatchError as e:
logging.debug('WATCH ERROR. Someone else is writing to the catalog. Retrying...')
continue
self.data[key]['xid:start'] = start_index
self.data[key]['xid:end'] = start_index + traj.n_frames
bench.mark('catalog')
# ---- POST PROCESSING
# Check for total data in downstream queue & Notify control manager to run
# 1. Det. what's still running
job_queue = slurm.currentqueue()
n_simjobs = -1 # -1 to exclude self
for j in job_queue:
if j['name'].startswith('sw'):
n_simjobs += 1
logging.info('SIM WORKER has detected %d other simulations running/pending.', len(job_queue))
remain_percent = n_simjobs / self.data['numresources']
# Fault Tolerance: ensure the pipeline persists
if n_simjobs == 0:
logging.info('I am the last simulation. Ensuring the controler executes.')
self.catalog.set('ctl:force', 1)
# Send downstream notification when less than 10% of jobs are still running
if not self.skip_notify and remain_percent < .1:
# Notify the control manager 'CM'
self.notify('ctl')
if USE_SHM:
shutil.rmtree(ramdisk)
shm_contents = os.listdir('/dev/shm')
logging.debug('Ramdisk contents (should be empty of DDC) : %s', str(shm_contents))
# For benchmarching:
bench.show()
stat.show()
# Return # of observations (frames) processed
return downstream_list
def bootstrap_lattice(catalog, num=10, build_new=False):
''' Bootstrap After TimeScape has run on source trajectory '''
home = os.getenv("HOME")
support = 1
cutoff = 8
start_coord = [
'de2586_315', 'de531_20', 'de3765_63', 'de3305_668', 'de1732_139'
]
dcdfile = lambda x: home + '/work/data/{0}.dcd'.format(x)
outloc = lambda x: home + '/work/jc/denovouniform1/{0}/{0}'.format(x)
traj_list = {}
basin_list = catalog.lrange('basin:list', 0, -1)
if len(basin_list) == 134:
logging.info('Basin Data already loaded!')
rms_delta_list = [
(i, np.sum(pickle.loads(catalog.get('basin:rmsdelta:' + b))))
for i, b in enumerate(basin_list)
]
else:
logging.info('Loading all bootstrap data to initialize...')
basin_list = []
rms_delta_list = []
pdb_file = home + '/work/data/alpha.pdb'
topo = md.load(pdb_file)
ref_alpha = md.load(home + '/work/' + catalog.get('pdb:ref:0'))
ref_alpha.atom_slice(ref_alpha.top.select_atom_indices('alpha'),
inplace=True)
res_rms_Kr = FEATURE_SET
for sc in start_coord:
dist_space = []
srcfile = outloc(sc) + '.dcd'
pdbfile = srcfile.replace('dcd', 'pdb')
logging.debug('LOADING TRAJ: %s', srcfile)
traj = md.load(srcfile, top=pdbfile)
traj_list[sc] = traj
alpha = traj.atom_slice(traj.top.select_atom_indices('alpha'))
logging.info('Grabbing TS data...')
W = TS.TimeScape.windows(outloc(sc) + '_transitions.log')
ts_traj = TS.TimeScapeParser(pdbfile,
outloc(sc),
sc,
dcd=srcfile,
traj=traj)
basins = ts_traj.load_basins()
logging.info("Processing distance space and residue RMS")
dsa = DR.distance_space(alpha)
resrmsd = 10 * np.array(
[LA.norm(i - ref_alpha.xyz[0], axis=1) for i in alpha.xyz])
basin_res_rms = np.zeros(shape=(len(ts_traj.basins),
alpha.n_atoms))
for i, (a, b) in enumerate(W):
dist_space.append(dsa[a:b].mean(0))
basin_res_rms[i] = np.median(resrmsd[a:b], axis=0)
basin_res_rms_delta = np.array(
[rms_delta(i) for i in basin_res_rms.T]).T
logging.debug('RMS LEN CHECK: %d =?= %d -- Updating RMS Delta',
len(basins), len(basin_res_rms_delta))
for i, basin in enumerate(basins):
pipe = catalog.pipeline()
bid = basin.id
# Store on Disk and in redis
jc_filename = os.path.join(settings.datadir,
'basin_%s.pdb' % bid)
logging.info('MIN for %s: Idx# %d to %s', bid, basin.mindex,
jc_filename)
minima_frame = traj.slice(
basin.mindex
) #md.load_frame(src_traj, basin.mindex, top=src_traj.replace('dcd', 'pdb'))
minima_frame.save_pdb(jc_filename)
basin_hash = basin.kv()
basin_hash['pdbfile'] = jc_filename
logging.info(
' Basin: %(id)s %(start)d - %(end)d Minima: %(mindex)d size=%(len)d'
% basin_hash)
pipe.rpush('basin:list', bid)
pipe.hmset('basin:%s' % bid, basin_hash)
pipe.set('basin:dmu:' + bid, pickle.dumps(dist_space[i]))
pipe.set('minima:%s' % bid, pickle.dumps(minima_frame))
# FOR RESIDUE RMSD
resrms_d = np.sum(basin_res_rms_delta[i][res_rms_Kr])
basin_hash['resrms_delta'] = resrms_d
rms_delta_list.append((len(basin_list), resrms_d))
basin_list.append(basin_hash)
pipe.set('basin:rmsdelta:' + bid,
pickle.dumps(basin_res_rms_delta[i]))
pipe.execute()
# FOR RESIDUE RMSD
# FOR SEED SAMPLING USING RMS_DELTA
# Note: skip the first basin
# Re-Construct the Lattice from
if build_new:
dist_space = 10 * np.array(dist_space)
cm = ds < cutoff
fs = lat.reduced_feature_set(cm, .115)
len(fs)
dr, cr = ds[:, fs], cm[:, fs]
mfis, lfis = lat.maxminer(cr, 1)
dlat, ik = lat.derived_lattice(mfis, dr, cr)
pickle.dump(mfis, open(home + '/work/data/denovo_mfis.p', 'wb'))
pickle.dump(lfis, open(home + '/work/data/denovo_lfis.p', 'wb'))
pickle.dump(ik, open(home + '/work/data/denovo_iset.p', 'wb'))
pickle.dump(dlat, open(home + '/work/data/denovo_dlat.p', 'wb'))
else:
logging.info('Loading Pre-Constructed Lattice Data')
dlat = pickle.load(open(home + '/work/data/denovo_dlat.p', 'rb'))
mfis = pickle.load(open(home + '/work/data/denovo_mfis.p', 'rb'))
lfis = pickle.load(open(home + '/work/data/denovo_lfis.p', 'rb'))
ik = pickle.load(open(home + '/work/data/denovo_iset.p', 'rb'))
with catalog.pipeline() as pipe:
pipe.set('lattice:max_fis', pickle.dumps(mfis))
pipe.set('lattice:low_fis', pickle.dumps(lfis))
pipe.set('lattice:dlat', pickle.dumps(dlat))
pipe.set('lattice:iset', pickle.dumps(ik))
pipe.execute()
# logging.info('Building Existing lattice object')
# lattice=lat.Lattice(ds, fs, cutoff, support)
# lattice.set_fis(max_fis, low_fis)
# lattice.set_dlat(dlat, Ik)
# sampler = LatticeSampler(lattice)
# Sample -- FOR USING LATTICE TO BOOTSTRAP
# cl,sc,el = lat.clusterlattice(dlat, cr, dr, ik, num_k=8, invert=True)
# cl_list = sorted(el, key=lambda x: len(x))
# TODO: Check if fan out > single item clusters
# start_indices = [clu[0][0] for clu in cl_list[:num]]
rms_delta_ranked = [
x[0] for x in sorted(rms_delta_list, key=lambda i: i[1], reverse=True)
]
start_indices = rms_delta_ranked[:num]
seedlist = [catalog.lindex('basin:list', i) for i in start_indices]
sim_init = {key: catalog.get(key) for key in settings.sim_params.keys()}
global_params = getSimParameters(sim_init, 'seed')
global_params['psf'] = home + '/work/jc/serial2/de0_0/de0_0.psf'
for seed in seedlist:
logging.debug('\nSeeding Job: %s ', seed)
basin = catalog.hgetall('basin:%s' % seed)
catalog.rpush('executed', seed)
# Generate new set of params/coords
jcID, config = generateFromBasin(basin)
# Update Additional JC Params and Decision History, as needed
config.update(global_params)
# Push to catalog
logging.info("New Simulation Job Created: %s", jcID)
for k, v in config.items():
logging.debug(" %s: %s", k, str(v))
catalog.rpush('jcqueue', jcID)
catalog.hmset(wrapKey('jc', jcID), config)
def load_historical_Expl(catalog):
""" Load all DEShaw data into basins for processing """
settings = systemsettings()
# idx_list = [0,1,2,3,4,20,23,24,30,32,34,40,41,42]
idx_list = [0, 34]
tlist = {k: 'tr%d.dcd' % k for k in idx_list}
seed_dir = os.path.join(settings.WORKDIR, 'seed')
seed_ts_ratio = 16 # TimeScape ran on 4ps frame rate (16x source)
# Load topology and anscillary data
# bpti = Protein(bpti, catalog, load=True)
pdb_file = os.path.join(seed_dir, 'coord.pdb')
topo = md.load(pdb_file)
pfilt = topo.top.select('protein')
logging.info('Topology loaded %s', topo)
# ID Side Chain pair atoms for each distance space calc
sc_pairs = side_chain_pairs(topo.atom_slice(pfilt))
logging.info('Identified side chains: %d', len(sc_pairs))
# Process all sorce trajectories
basin_list = []
C_T, mu_T, sigma_T = [], [], []
for idx in idx_list:
logging.info('Procesing Seed index: %d', idx)
# Load SRC seed trajetory & calc distance space -- TODO: make this optional
tfile = os.path.join(seed_dir, tlist[idx])
traj = md.load(tfile, top=topo)
traj.superpose(topo)
ds = datareduce.distance_space(traj, pairs=sc_pairs)
# Push to Catalog
file_idx = catalog.rpush('xid:filelist', tfile) - 1
start_index = catalog.llen('xid:reference')
# TODO: Do I still need to index every frame??????
catalog.rpush('xid:reference',
*[(file_idx, x) for x in range(traj.n_frames)])
# Process Trajectory as basins
logging.info(" Seed Loaded. Loading TimeScape Data...")
seed_name = 'tr%d' % idx
ts_data_path = os.path.join(seed_dir, 'TEST', seed_name)
ts_traj = TimeScapeParser(pdb_file,
ts_data_path,
seed_name,
dcd=tfile,
traj=traj)
basin_list = ts_traj.load_basins(frame_ratio=seed_ts_ratio)
corr_mat = ts_traj.correlation_matrix()
for i, basin in enumerate(ts_traj.basins):
a, b = basin.start, basin.end
bid = basin.id
if a > traj.n_frames:
logging.info(
'Finished processing all basins for this Trajectory!')
break
# Store on Disk and in redis
jc_filename = os.path.join(settings.datadir, 'basin_%s.pdb' % bid)
minima_frame = md.load_frame(
tfile, basin.mindex, top=topo) if traj is None else traj.slice(
basin.mindex)
minima_frame.save_pdb(jc_filename)
C_T.append(np.mean(corr_mat[a:b], axis=0))
mu_T.append(np.mean(ds[a:b], axis=0))
sigma_T.append(np.std(ds[a:b], axis=0))
basin_hash = basin.kv()
basin_hash['pdbfile'] = jc_filename
basin_hash['dsidx'] = i
logging.info(
' Basin: %(id)s %(start)d - %(end)d Minima: %(mindex)d size=%(len)d'
% basin_hash)
pipe = catalog.pipeline()
pipe.rpush('basin:list', bid)
pipe.hmset('basin:%s' % bid, basin_hash)
pipe.set('minima:%s' % bid, pickle.dumps(minima_frame))
pipe.execute()
catalog.set('basin:processed', len(ts_traj.basins))
catalog.storeNPArray(np.array(C_T), 'corr_vector')
catalog.storeNPArray(np.array(mu_T), 'dspace_mu')