def dijkstra(G, s):
"""no negative edge"""
init_single_source(G, s)
S = {}
Q = [] # min-priority queue
# added to conditionally push node to Q
seen = {}
c = count()
pop = heappop
push = heappush
push(Q, (s.d, next(c), s))
while Q:
dist, _, u = pop(Q)
if u in S:
continue
# shortest path for 'u' are calculated over before checking adj
S[u] = dist
for v in G.adj[u]:
w = G.adj[u][v].get('w', 1)
uv_dist = dist + w
if v not in S and (v not in seen or uv_dist < seen[v]):
relax(u, v, w)
push(Q, (v.d, next(c), v))
return S
def dag_shortest_path(G, s):
"""no cycle"""
vertices_sorted = topological_sort(G)
init_single_source(G, s)
for u in vertices_sorted:
for v in G.adj[u]:
relax(u, v, G.adj[u][v]['w'])
def dijkstra(self, g, s):
initialize_single_source(g, s)
S = []
Q = [v for v in g.get_vertices()]
while len(Q):
u = extract_min(Q)
u.set_visitado()
S.append(u)
for v in u.get_vertices_adjacentes():
if v.get_visitado():
continue
relax(u, v)
def bellman_ford(self, g, s):
initialize_single_source(g, s)
for u in g.get_vertices():
for u, v in g.get_arestas():
if v.get_visitado():
continue
relax(u, v)
for u, v in g.get_arestas():
if v.get_distancia() > u.get_distancia() + u.get_peso(v):
return False
return True
def simulate(self,
sim_index: int,
seed: int = 9001) -> Tuple[int, float, float]:
"""
Run full MCMC simulation from start_temp to end_temp.
Be sure to save the best (lowest-energy) structure, so you can access it after.
It is also a good idea to track certain variables during the simulation (temp, energy, and more).
:param sim_index: simulation index
:param seed: int
:return: log information
"""
random.seed(seed)
log = {"temperature": [], "energy": []}
while True:
if self.step():
# store best pdb
sim_path, log_folder_path, sim_index = self.storeSim(
self.protein, log, sim_index)
# calculate relaxed, cd to the folder!!
cur_dir = os.getcwd()
os.chdir(sim_path)
protein, rmsd, score = utils.relax("best.pdb", "target.pdb")
os.chdir(cur_dir)
break
else:
# keep track of log
log['temperature'].append(self.temp)
log['energy'].append(self.compute_energy())
return sim_index, score, rmsd
def bellman_ford(G, s):
"""no negative cycle"""
init_single_source(G, s)
# for every node in G
for u in G.node[:-1]:
# for every edge in G
for u, vs in G.adj.items():
for v in vs:
w = G.adj[u][v]['w']
relax(u, v, w)
for u in G.node:
for v in G.adj[u]:
w = G.adj[u][v]['w']
if v.d > u.d + w: # when w is negative
print("Error", u, v, w)
return False
return True
def main():
#Read in arguments from the command line
parser = argparse.ArgumentParser(
description='Performs ab initio protein structure prediction')
parser.add_argument('--fasta',
help='name of the fasta file containing the sequence')
parser.add_argument(
'--logdir', help='directory to save all log files',
default='.') #changed so does not have the / afterwards
parser.add_argument('--nsims',
type=int,
help='number of simulations',
default=1)
parser.add_argument('--nfrags',
type=int,
help='number of fragments to sample at each iteration',
default=3)
parser.add_argument('--anneal_rate',
type=float,
help='temperature annealing parameter',
default=0.999)
args = parser.parse_args()
#load sequence from fasta file
my_file = open(args.fasta)
my_sequence = my_file.readlines()
my_sequence = my_sequence[1].strip()
my_file.close()
#initialize protein into extended conformation
my_protein = Protein(sequence=my_sequence)
#set up FragmentSets for later use
protein_name = args.fasta[:-6]
fragfile_9 = protein_name + '_9mers.frag'
rmsdfile_9 = protein_name + '_9mers.rmsd'
fragment_set_9mer = FragmentSet(fragfile_9, rmsdfile_9)
fragfile_3 = protein_name + '_3mers.frag'
rmsdfile_3 = protein_name + '_3mers.rmsd'
fragment_set_3mer = FragmentSet(fragfile_3, rmsdfile_3)
#create lists for saving certain information about each simulation
sim_numbers = []
sim_rmsds = []
sim_energies = []
sim_proteins = []
#loop through 10 sims
for sim_number in range(1, args.nsims + 1):
#create FragmentSampler object and run coarse simulation with 9mers
coarse_sampler = MCMCSampler(my_protein,
fragment_set_9mer,
nmers=9,
start_temp=100,
end_temp=1,
nfrags=3,
anneal_rate=args.anneal_rate)
coarse_result = coarse_sampler.simulate()
#create FragmentSampler object and run refinement simulation with 3mers
fine_sampler = MCMCSampler(coarse_sampler.best_protein,
fragment_set_3mer,
nmers=3,
start_temp=1,
end_temp=0.1,
nfrags=3,
anneal_rate=args.anneal_rate)
fine_result = fine_sampler.simulate()
#save the final structure from the simulation
fine_sampler.best_protein.save_pdb(filename='best.pdb')
#save a log_file from the simulation
total_result = pd.concat([
coarse_result, fine_result
]) #might need some fixing depending on what the quiz asks for
total_result.to_csv(
(args.logdir + '/sim_' + str(sim_number) + '_log.txt'), sep='\t')
#relax the simulated structure
relaxed_protein, final_rmsd, final_energy = utils.relax(
pdb='best.pdb', native=args.fasta.split(sep='.')[0] + '.pdb')
#log some parameters...sim_number, energy, rmsd
sim_numbers.append(sim_number)
sim_rmsds.append(final_rmsd)
sim_energies.append(final_energy)
sim_proteins.append(relaxed_protein)
#return sim log
sim_log = pd.DataFrame(list(zip(sim_numbers, sim_energies, sim_rmsds)),
columns=['sim_number', 'energy', 'rmsd'])
sim_log.to_csv(
(args.logdir + '/simulation_summary.txt'), sep='\t', index=False
) #may need to be tweaked, i.e. removing indices in first column