Skip to main content
  • Home
  • Development
  • Documentation
  • Donate
  • Operational login
  • Browse the archive

swh logo
SoftwareHeritage
Software
Heritage
Archive
Features
  • Search

  • Downloads

  • Save code now

  • Add forge now

  • Help

Revision 4816f2a300d7bbae1084b1940959ea35973104e1 authored by Mingzhi Lin on 23 August 2017, 14:12:59 UTC, committed by Mingzhi Lin on 23 August 2017, 14:12:59 UTC
add nucl_cov_test
1 parent 4453664
  • Files
  • Changes
  • ffcddc8
  • /
  • nucl_cov.go
Raw File Download
Permalinks

To reference or cite the objects present in the Software Heritage archive, permalinks based on SoftWare Hash IDentifiers (SWHIDs) must be used.
Select below a type of object currently browsed in order to display its associated SWHID and permalink.

  • revision
  • directory
  • content
revision badge
swh:1:rev:4816f2a300d7bbae1084b1940959ea35973104e1
directory badge Iframe embedding
swh:1:dir:ffcddc89ba2c0f13614ac49f75a9bda0bb0d8a58
content badge Iframe embedding
swh:1:cnt:f761b5e21ed9fda18dcdfcb720b97ddda9b5c776
Citations

This interface enables to generate software citations, provided that the root directory of browsed objects contains a citation.cff or codemeta.json file.
Select below a type of object currently browsed in order to generate citations for them.

  • revision
  • directory
  • content
Generate software citation in BibTex format (requires biblatex-software package)
Generating citation ...
Generate software citation in BibTex format (requires biblatex-software package)
Generating citation ...
Generate software citation in BibTex format (requires biblatex-software package)
Generating citation ...
nucl_cov.go
package mcorr

import (
	"bytes"
	"fmt"
)

// NuclCov contains covariance of nucleotide acid in a DNA sequence.
type NuclCov struct {
	Doublets []int
	Alphabet []byte
}

// NewNuclCov return a NuclCov given the alphabet.
func NewNuclCov(alphabet []byte) *NuclCov {
	sizeOfAlphabet := len(alphabet)
	nc := NuclCov{Alphabet: alphabet}
	nc.Doublets = make([]int, sizeOfAlphabet*sizeOfAlphabet)
	return &nc
}

// Add insert a pair of nucliotide acids.
// It returns error when the nucliotide acid is not in the alphabet.
func (nc *NuclCov) Add(a, b byte) error {
	indexA := bytes.IndexByte(nc.Alphabet, a)
	indexB := bytes.IndexByte(nc.Alphabet, b)
	sizeOfAlphabet := len(nc.Alphabet)
	if indexA >= 0 && indexB >= 0 {
		nc.Doublets[indexA*sizeOfAlphabet+indexB]++
		return nil
	}

	var err error
	if indexA < 0 && indexB < 0 {
		err = fmt.Errorf("%c and %c are not in Alphabet: %s", a, b, string(nc.Alphabet))
	} else if indexA < 0 {
		err = fmt.Errorf("%c is not in Alphabet: %s", a, string(nc.Alphabet))
	} else {
		err = fmt.Errorf("%c is not in Alphabet: %s", b, string(nc.Alphabet))
	}

	return err
}

// Count returns the total number of pairs.
func (nc *NuclCov) Count() int {
	n := 0
	for _, a := range nc.Doublets {
		n += a
	}
	return n
}

// XBar return the count of 1 at the first position
func (nc *NuclCov) XBar() (x float64, n int) {
	sizeOfAlphabet := len(nc.Alphabet)
	for i := 0; i < len(nc.Doublets); i++ {
		if nc.Doublets[i] > 0 {
			for j := i + 1; j < len(nc.Doublets); j++ {
				if nc.Doublets[j] > 0 {
					c := float64(nc.Doublets[i] * nc.Doublets[j])

					if i/sizeOfAlphabet != j/sizeOfAlphabet {
						x += c
					}
					n += nc.Doublets[i] * nc.Doublets[j]
				}
			}
			n += nc.Doublets[i] * (nc.Doublets[i] - 1) / 2
		}
	}
	return
}

// YBar return the count of 1 at the first position
func (nc *NuclCov) YBar() (x float64, n int) {
	sizeOfAlphabet := len(nc.Alphabet)
	for i := 0; i < len(nc.Doublets); i++ {
		if nc.Doublets[i] > 0 {
			for j := i + 1; j < len(nc.Doublets); j++ {
				if nc.Doublets[j] > 0 {
					c := float64(nc.Doublets[i] * nc.Doublets[j])
					if i%sizeOfAlphabet != j%sizeOfAlphabet {
						x += c
					}
					n += nc.Doublets[i] * nc.Doublets[j]
				}
			}
			n += nc.Doublets[i] * (nc.Doublets[i] - 1) / 2
		}
	}
	return
}

// P00 returns the covariance.
func (nc *NuclCov) P00() (xy float64, n int) {
	for i := 0; i < len(nc.Doublets); i++ {
		if nc.Doublets[i] > 0 {
			for j := i + 1; j < len(nc.Doublets); j++ {
				if nc.Doublets[j] > 0 {
					n += nc.Doublets[i] * nc.Doublets[j]
				}
			}
			n += nc.Doublets[i] * (nc.Doublets[i] - 1) / 2
			xy += float64(nc.Doublets[i] * (nc.Doublets[i] - 1) / 2)
		}
	}
	return
}

// P01 returns the covariance.
func (nc *NuclCov) P01() (xy float64, n int) {
	sizeOfAlphabet := len(nc.Alphabet)
	for i := 0; i < len(nc.Doublets); i++ {
		if nc.Doublets[i] > 0 {
			for j := i + 1; j < len(nc.Doublets); j++ {
				if nc.Doublets[j] > 0 {
					c := float64(nc.Doublets[i] * nc.Doublets[j])
					if i%sizeOfAlphabet != j%sizeOfAlphabet {
						xy += c
					}
					n += nc.Doublets[i] * nc.Doublets[j]
				}
			}
			n += nc.Doublets[i] * (nc.Doublets[i] - 1) / 2
		}
	}
	return
}

// P10 returns the covariance.
func (nc *NuclCov) P10() (xy float64, n int) {
	sizeOfAlphabet := len(nc.Alphabet)
	for i := 0; i < len(nc.Doublets); i++ {
		if nc.Doublets[i] > 0 {
			for j := i + 1; j < len(nc.Doublets); j++ {
				if nc.Doublets[j] > 0 {
					c := float64(nc.Doublets[i] * nc.Doublets[j])
					if i/sizeOfAlphabet != j/sizeOfAlphabet {
						xy += c
					}
					n += nc.Doublets[i] * nc.Doublets[j]
				}
			}
			n += nc.Doublets[i] * (nc.Doublets[i] - 1) / 2
		}
	}
	return
}

// P11 returns the covariance.
func (nc *NuclCov) P11() (xy float64, n int) {
	sizeOfAlphabet := len(nc.Alphabet)
	for i := 0; i < len(nc.Doublets); i++ {
		if nc.Doublets[i] > 0 {
			for j := i + 1; j < len(nc.Doublets); j++ {
				if nc.Doublets[j] > 0 {
					c := float64(nc.Doublets[i] * nc.Doublets[j])
					if i%sizeOfAlphabet != j%sizeOfAlphabet && i/sizeOfAlphabet != j/sizeOfAlphabet {
						xy += c
					}

					n += nc.Doublets[i] * nc.Doublets[j]
				}
			}
			n += nc.Doublets[i] * (nc.Doublets[i] - 1) / 2
		}
	}
	return
}

// MateP11 calculate covariance between two clusters.
func (nc *NuclCov) MateP11(nc2 *NuclCov) (xy float64, n int) {
	sizeOfAlphabet := len(nc.Alphabet)
	for i := 0; i < len(nc.Doublets); i++ {
		if nc.Doublets[i] > 0 {
			for j := 0; j < len(nc2.Doublets); j++ {
				if i != j && nc2.Doublets[j] > 0 {
					c := float64(nc.Doublets[i] * nc2.Doublets[j])
					if i%sizeOfAlphabet != j%sizeOfAlphabet && i/sizeOfAlphabet != j/sizeOfAlphabet {
						xy += c
					}
				}
			}
		}
	}
	n1 := 0
	n2 := 0
	for i := 0; i < len(nc.Doublets); i++ {
		n1 += nc.Doublets[i]
		n2 += nc2.Doublets[i]
	}
	n = n1 * n2
	return
}

// MateP00 calculate covariance between two clusters.
func (nc *NuclCov) MateP00(nc2 *NuclCov) (xy float64, n int) {
	n1, n2 := 0, 0
	for i := 0; i < len(nc.Doublets); i++ {
		xy += float64(nc.Doublets[i] * nc2.Doublets[i])
		n1 += nc.Doublets[i]
		n2 += nc2.Doublets[i]
	}
	n = n1 * n2
	return
}

// Append another NuclCov.
func (nc *NuclCov) Append(nc2 *NuclCov) error {
	// Check alphabet
	diffAlphabetError := fmt.Errorf("Different alphbet %s, %s", string(nc.Alphabet), string(nc2.Alphabet))
	if len(nc.Alphabet) != len(nc2.Alphabet) {
		return diffAlphabetError
	}
	for i, a := range nc.Alphabet {
		b := nc2.Alphabet[i]
		if a != b {
			return diffAlphabetError
		}
	}

	for i := 0; i < len(nc.Doublets); i++ {
		nc.Doublets[i] += nc2.Doublets[i]
	}

	return nil
}
The diff you're trying to view is too large. Only the first 1000 changed files have been loaded.
Showing with 0 additions and 0 deletions (0 / 0 diffs computed)
swh spinner

Computing file changes ...

Software Heritage — Copyright (C) 2015–2025, The Software Heritage developers. License: GNU AGPLv3+.
The source code of Software Heritage itself is available on our development forge.
The source code files archived by Software Heritage are available under their own copyright and licenses.
Terms of use: Archive access, API— Contact— JavaScript license information— Web API

back to top