func getPdbChain(fp string) *pdb.Chain {
b := path.Base(fp)
if !strings.HasSuffix(b, ".fmap") {
util.Fatalf("Expected file named 'something.fmap' but got '%s'.", b)
}
idAndChain := b[0 : len(b)-5]
if len(idAndChain) != 5 {
util.Fatalf("Expected 4-letter PDB id concatenated with 1-letter "+
"chain identifier, but got '%s' instead.", idAndChain)
}
pdbName := idAndChain[0:4]
chainId := idAndChain[4]
pdbCat := idAndChain[1:3]
pdbFile := fmt.Sprintf("pdb%s.ent.gz", pdbName)
pdbPath := path.Join(util.FlagPdbDir, pdbCat, pdbFile)
entry := util.PDBRead(pdbPath)
chain := entry.Chain(chainId)
if chain == nil {
util.Fatalf("Could not find chain '%c' in PDB entry '%s'.",
chainId, pdbPath)
}
return chain
}
func main() {
var f io.Reader
var err error
f = util.OpenFile(flag.Arg(0))
if strings.HasSuffix(flag.Arg(0), ".gz") {
f, err = gzip.NewReader(f)
util.Assert(err)
}
cifEntry, err := pdbx.Read(f)
util.Assert(err, "Could not read PDBx/mmCIF file")
fasEntries := make([]seq.Sequence, 0, 5)
for _, ent := range cifEntry.Entities {
for _, chain := range ent.Chains {
if !isChainUsable(chain) || len(ent.Seq) == 0 {
continue
}
fasEntry := seq.Sequence{
Name: chainHeader(chain),
Residues: ent.Seq,
}
fasEntries = append(fasEntries, fasEntry)
}
}
if len(fasEntries) == 0 {
util.Fatalf("Could not find any chains with amino acids.")
}
var fasOut io.Writer
if flag.NArg() == 1 {
fasOut = os.Stdout
} else {
if len(flagSplit) > 0 {
util.Fatalf("The '--split' option is incompatible with a single " +
"output file.")
}
fasOut = util.CreateFile(util.Arg(1))
}
if len(flagSplit) == 0 {
util.Assert(fasta.NewWriter(fasOut).WriteAll(fasEntries),
"Could not write FASTA file '%s'", fasOut)
} else {
for _, entry := range fasEntries {
fp := path.Join(flagSplit, fmt.Sprintf("%s.fasta", entry.Name))
out := util.CreateFile(fp)
w := fasta.NewWriter(out)
util.Assert(w.Write(entry), "Could not write to '%s'", fp)
util.Assert(w.Flush(), "Could not write to '%s'", fp)
}
}
}
func mkPaired(c *command) {
c.assertNArg(2)
in := util.Library(c.flags.Arg(0))
outPath := c.flags.Arg(1)
util.AssertOverwritable(outPath, flagOverwrite)
if _, ok := in.(fragbag.WeightedLibrary); ok {
util.Fatalf("%s is a weighted library (not allowed)", in.Name())
}
name := fmt.Sprintf("paired-%s", in.Name())
if fragbag.IsStructure(in) {
var pairs [][]structure.Coords
lib := in.(fragbag.StructureLibrary)
nfrags := lib.Size()
for i := 0; i < nfrags; i++ {
for j := 0; j < nfrags; j++ {
if i == j {
continue
}
f1, f2 := lib.Atoms(i), lib.Atoms(j)
pairs = append(pairs, append(f1, f2...))
}
}
pairLib, err := fragbag.NewStructureAtoms(name, pairs)
util.Assert(err)
fragbag.Save(util.CreateFile(outPath), pairLib)
} else if strings.Contains(in.Tag(), "hmm") {
var pairs []*seq.HMM
lib := in.(fragbag.SequenceLibrary)
nfrags := lib.Size()
for i := 0; i < nfrags; i++ {
for j := 0; j < nfrags; j++ {
if i == j {
continue
}
f1, f2 := lib.Fragment(i).(*seq.HMM), lib.Fragment(j).(*seq.HMM)
pairs = append(pairs, seq.HMMCat(f1, f2))
}
}
pairLib, err := fragbag.NewSequenceHMM(name, pairs)
util.Assert(err)
fragbag.Save(util.CreateFile(outPath), pairLib)
} else if strings.Contains(in.Tag(), "profile") {
util.Fatalf("Sequence profiles not implemented.")
} else {
util.Fatalf("Unrecognized fragment library: %s", in.Tag())
}
}
func main() {
entry := util.PDBRead(flag.Arg(0))
if len(flagChain) > 0 {
if len(flagChain) != 1 {
util.Fatalf("Chain identifiers must be a single character.")
}
chain := entry.Chain(flagChain[0])
if chain == nil {
util.Fatalf("Could not find chain '%c' in PDB entry '%s'.",
chain.Ident, entry.Path)
}
showMapping(chain, chain.SequenceAtoms())
} else {
for _, chain := range entry.Chains {
showMapping(chain, chain.SequenceAtoms())
}
}
}
func coords(num int, atomRecords []byte) []structure.Coords {
r := bytes.NewReader(atomRecords)
name := fmt.Sprintf("fragment %d", num)
entry, err := pdb.Read(r, name)
util.Assert(err, "Fragment contents could not be read in PDB format")
atoms := entry.OneChain().CaAtoms()
if len(atoms) == 0 {
util.Fatalf("Fragment %d has no ATOM coordinates.", num)
}
return atoms
}
func ioFromFile(fpath, force string) msaIO {
var fmt string
if len(force) > 0 {
fmt = force
} else {
var ok bool
ext := path.Ext(fpath)
if len(ext) > 0 {
ext = ext[1:]
}
fmt, ok = extToFmt[ext]
if !ok {
util.Fatalf("Could not detect format from extension '%s'.", ext)
}
}
io, ok := fmtToIO[fmt]
if !ok {
util.Fatalf("BUG: Could not find converters for format '%s'.", fmt)
}
return io
}
func main() {
lib = util.StructureLibrary(util.Arg(0))
pdbEntry := util.PDBRead(util.Arg(1))
if util.NArg() == 2 {
for _, chain := range pdbEntry.Chains {
atoms := chain.CaAtoms()
bestFragsForRegion(chain, atoms, 0, len(atoms))
}
} else {
chainId := util.Arg(2)
chain := pdbEntry.Chain(chainId[0])
if chain == nil || !chain.IsProtein() {
util.Fatalf("Could not find protein chain with id '%c'.", chainId)
}
atoms := chain.CaAtoms()
if util.NArg() == 3 {
bestFragsForRegion(chain, atoms, 0, len(atoms))
} else {
if util.NArg() != 5 {
log.Println("Both a start and end must be provided.")
util.Usage()
}
s, e := util.Arg(3), util.Arg(4)
sn, en := util.ParseInt(s)-1, util.ParseInt(e)
if en-sn < lib.FragmentSize() {
util.Fatalf("The range [%s, %s] specifies %d alpha-carbon "+
"atoms while at least %d alpha-carbon atoms are required "+
"for the given fragment library.",
s, e, en-sn, lib.FragmentSize())
}
bestFragsForRegion(chain, atoms, sn, en)
}
}
}
func search(c *command) {
c.assertLeastNArg(2)
// Some search options don't translate directly to command line parameters
// specified by the flag package.
if flagSearchDesc {
flagSearchOpts.Order = bowdb.OrderDesc
}
switch flagSearchSort {
case "cosine":
flagSearchOpts.SortBy = bowdb.SortByCosine
case "euclid":
flagSearchOpts.SortBy = bowdb.SortByEuclid
default:
util.Fatalf("Unknown sort field '%s'.", flagSearchSort)
}
db := util.OpenBowDB(c.flags.Arg(0))
bowPaths := c.flags.Args()[1:]
_, err := db.ReadAll()
util.Assert(err, "Could not read BOW database entries")
// always hide the progress bar here.
bows := util.ProcessBowers(bowPaths, db.Lib, false, flagCpu, true)
out, outDone := outputter()
// launch goroutines to search queries in parallel
wgSearch := new(sync.WaitGroup)
for i := 0; i < flagCpu; i++ {
wgSearch.Add(1)
go func() {
defer wgSearch.Done()
for b := range bows {
sr := db.Search(flagSearchOpts, b)
out <- searchResult{b, sr}
}
}()
}
wgSearch.Wait()
close(out)
<-outDone
util.Assert(db.Close())
}
func init() {
flag.Float64Var(&flagThreshold, "threshold", flagThreshold,
"Set the distance threshold to use when computing AUC.")
util.FlagUse("cpu", "cpuprof")
util.FlagParse(
"cath-domain-labels best-of-all-matrix"+
"(bowdb | matrix-file) out-file "+
"[ (bowdb | matrix-file) out-file ... ]",
"Computes the AUC of each aligner matrix (or BOW database) given\n"+
"with respect to the 'best-of-all' matrix given. Each AUC is\n"+
"written to a separate out-file. The sizes of all matrices must\n"+
"be exactly equivalent.\n"+
"Files are interpreted as BOW databases if they have a '.bowdb'\n"+
"file extension.")
util.AssertLeastNArg(4)
if util.NArg()%2 != 0 {
util.Fatalf("There must be an out file for each matrix or bowdb file.")
}
}
func main() {
libPath := util.Arg(0)
chain := util.Arg(1)
pdbEntryPath := util.Arg(2)
bowOut := util.Arg(3)
lib := util.StructureLibrary(libPath)
entry := util.PDBRead(pdbEntryPath)
thechain := entry.Chain(chain[0])
if thechain == nil || !thechain.IsProtein() {
util.Fatalf("Could not find chain with identifier '%c'.", chain[0])
}
bow := bow.BowerFromChain(thechain).StructureBow(lib)
if bowOut == "--" {
fmt.Println(bow)
} else {
util.BowWrite(util.CreateFile(bowOut), bow)
}
}
func recordToDist(record []string) pair {
namePieces := strings.SplitN(record[0], ".ent_", 2)
if len(namePieces) != 2 {
util.Fatalf("Invalid alignment pair: '%s'.", record[0])
}
p1, p2 := namePieces[0], namePieces[1]
p2 = p2[0 : len(p2)-5]
rf := func(i int) float64 { return readFloat(record[i]) }
corelen, rmsd := rf(1), rf(2)
l1, l2 := rf(7), rf(8)
coreval := (2.0 * corelen) / (l1 + l2)
dist := -6.04979701*(rmsd-coreval*corelen*0.155+1.6018) + 1000
dist = 1.0 / dist
dist *= 100.0
if p1 < p2 {
return pair{[2]string{p1, p2}, dist}
}
return pair{[2]string{p2, p1}, dist}
}
func outputter() (chan searchResult, chan struct{}) {
out := make(chan searchResult)
done := make(chan struct{})
go func() {
if flagSearchOutFmt == "csv" {
fmt.Printf("QueryID\tHitID\tCosine\tEuclid\n")
}
first := true
for sr := range out {
switch flagSearchOutFmt {
case "plain":
outputPlain(sr, first)
case "csv":
outputCsv(sr, first)
default:
util.Fatalf("Invalid output format '%s'.", flagSearchOutFmt)
}
first = false
}
done <- struct{}{}
}()
return out, done
}
func main() {
if len(util.FlagCpuProf) > 0 {
f := util.CreateFile(util.FlagCpuProf)
pprof.StartCPUProfile(f)
defer f.Close()
defer pprof.StopCPUProfile()
}
// Read all CATH domains, the best-of-all matrix, and the matrix for
// each aligner.
domains := readDomains(util.Arg(0))
boa := readMatrix(domains, util.Arg(1))
aligners := make([]aligner, 0)
flibs := make([]flib, 0)
for i := 2; i < util.NArg(); i += 2 {
fpath := util.Arg(i)
if path.Ext(fpath) == ".bowdb" {
db := util.OpenBowDB(fpath)
records, err := db.ReadAll()
util.Assert(err)
bowed := make([]bow.Bowed, domains.in.Len())
for _, b := range records {
if !domains.in.Exists(b.Id) {
util.Fatalf("Found ID in bowdb that isn't in the list "+
"of CATH domains provided: %s", b.Id)
}
bowed[domains.in.Atom(b.Id)] = b
}
flibs = append(flibs, flib{db, bowed, util.Arg(i + 1)})
} else {
aligners = append(aligners, aligner{
readMatrix(domains, fpath),
util.Arg(i + 1),
})
}
}
// Now remove CATH domains that don't have a corresponding structure file.
// We don't do this initially since the matrix files are indexed with
// respect to all CATH domains (includings ones without structure).
// This is an artifact of the fact that the matrices were generated with
// a very old version of CATH.
domains.removeOldDomains()
if a := matrixAuc(domains, boa, boa, flagThreshold); a != 1.0 {
util.Fatalf("Something is wrong. The AUC of the best-of-all matrix "+
"with respect to itself is %f, but it should be 1.0.", a)
}
if len(aligners) > 0 {
fmt.Println("Computing AUC for aligners...")
writeAuc := func(aligner aligner) struct{} {
w := util.CreateFile(aligner.outpath)
a := matrixAuc(domains, boa, aligner.matrix, flagThreshold)
fmt.Fprintf(w, "%f\n", a)
return struct{}{}
}
fun.ParMap(writeAuc, aligners)
}
if len(flibs) > 0 {
fmt.Println("Computing AUC for bowdbs...")
writeAuc := func(flib flib) struct{} {
w := util.CreateFile(flib.outpath)
a := flibAuc(domains, boa, flib, flagThreshold)
fmt.Fprintf(w, "%f\n", a)
return struct{}{}
}
fun.ParMap(writeAuc, flibs)
}
}
func main() {
pdbEntry := util.PDBRead(flag.Arg(0))
fasEntries := make([]seq.Sequence, 0, 5)
if !flagSeparateChains {
var fasEntry seq.Sequence
if len(pdbEntry.Chains) == 1 {
fasEntry.Name = chainHeader(pdbEntry.OneChain())
} else {
fasEntry.Name = fmt.Sprintf("%s", strings.ToLower(pdbEntry.IdCode))
}
seq := make([]seq.Residue, 0, 100)
for _, chain := range pdbEntry.Chains {
if isChainUsable(chain) {
seq = append(seq, chain.Sequence...)
}
}
fasEntry.Residues = seq
if len(fasEntry.Residues) == 0 {
util.Fatalf("Could not find any amino acids.")
}
fasEntries = append(fasEntries, fasEntry)
} else {
for _, chain := range pdbEntry.Chains {
if !isChainUsable(chain) {
continue
}
fasEntry := seq.Sequence{
Name: chainHeader(chain),
Residues: chain.Sequence,
}
fasEntries = append(fasEntries, fasEntry)
}
}
if len(fasEntries) == 0 {
util.Fatalf("Could not find any chains with amino acids.")
}
var fasOut io.Writer
if flag.NArg() == 1 {
fasOut = os.Stdout
} else {
if len(flagSplit) > 0 {
util.Fatalf("The '--split' option is incompatible with a single " +
"output file.")
}
fasOut = util.CreateFile(util.Arg(1))
}
if len(flagSplit) == 0 {
util.Assert(fasta.NewWriter(fasOut).WriteAll(fasEntries),
"Could not write FASTA file '%s'", fasOut)
} else {
for _, entry := range fasEntries {
fp := path.Join(flagSplit, fmt.Sprintf("%s.fasta", entry.Name))
out := util.CreateFile(fp)
w := fasta.NewWriter(out)
util.Assert(w.Write(entry), "Could not write to '%s'", fp)
util.Assert(w.Flush(), "Could not write to '%s'", fp)
}
}
}
func main() {
fmapPath := util.Arg(0)
fmap := util.FmapRead(fmapPath)
qchain := getPdbChain(fmapPath)
stats := newSequenceStats(qchain.Sequence)
total, trueps := 0, 0
qcorrupt, tcorrupt := 0, 0
for _, frags := range fmap.Segments {
for _, frag := range frags.Frags {
hit := frag.Hit
if frag.IsCorrupt() {
tcorrupt += 1
stats.incTCorrupt(hit)
continue
}
qatoms := qchain.SequenceCaAtomSlice(hit.QueryStart-1, hit.QueryEnd)
if qatoms == nil {
qcorrupt += 1
stats.incQCorrupt(hit)
continue
}
if len(qatoms) != len(frag.CaAtoms) {
util.Fatalf("Uncomparable lengths. Query is (%d, %d) while "+
"template is (%d, %d). Length of query CaAtoms: %d, "+
"length of template CaAtoms: %d",
hit.QueryStart, hit.QueryEnd,
hit.TemplateStart, hit.TemplateEnd,
len(qatoms), len(frag.CaAtoms))
}
if structure.RMSD(qatoms, frag.CaAtoms) <= flagRmsd {
trueps += 1
stats.incTruePs(hit)
}
total += 1
stats.incTotal(hit)
}
}
coveredResidues := 0
for _, resStats := range stats {
if resStats.trueps >= 1 {
coveredResidues += 1
}
}
coverage := float64(coveredResidues) / float64(len(qchain.Sequence))
fmt.Printf("RMSDThreshold: %f\n", flagRmsd)
fmt.Printf("TotalFragments: %d\n", total)
fmt.Printf("TruePositives: %d\n", trueps)
fmt.Printf("Precision: %f\n", float64(trueps)/float64(total))
fmt.Printf("CorruptQuery: %d\n", qcorrupt)
fmt.Printf("CorruptTemplate: %d\n", tcorrupt)
fmt.Printf("TotalResidues: %d\n", len(qchain.Sequence))
fmt.Printf("CoveredResidues: %d\n", coveredResidues)
fmt.Printf("Coverage: %f\n", coverage)
}
