1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155

use utils::*;

pub struct SaPartFuncMats {
  pub sa_forward_part_func_mat_4_char_align: PartFuncMat,
  pub sa_forward_part_func_mat_4_gap_1: PartFuncMat,
  pub sa_forward_part_func_mat_4_gap_2: PartFuncMat,
  pub sa_backward_part_func_mat_4_char_align: PartFuncMat,
  pub sa_backward_part_func_mat_4_gap_1: PartFuncMat,
  pub sa_backward_part_func_mat_4_gap_2: PartFuncMat,
}

impl SaPartFuncMats {
  fn new(slp: &(usize, usize)) -> SaPartFuncMats {
    let zero_mat = vec![vec![0.; slp.1 + 2]; slp.0 + 2];
    SaPartFuncMats {
      sa_forward_part_func_mat_4_char_align: zero_mat.clone(),
      sa_forward_part_func_mat_4_gap_1: zero_mat.clone(),
      sa_forward_part_func_mat_4_gap_2: zero_mat.clone(),
      sa_backward_part_func_mat_4_char_align: zero_mat.clone(),
      sa_backward_part_func_mat_4_gap_1: zero_mat.clone(),
      sa_backward_part_func_mat_4_gap_2: zero_mat,
    }
  }
}

impl SaScoringParams {
  pub fn new(ca_sm: &CaScoreMat, opening_gap_penalty: SaScore, extending_gap_penalty: SaScore) -> SaScoringParams {
    SaScoringParams {
      ca_sm: ca_sm.clone(),
      opening_gap_penalty: opening_gap_penalty,
      extending_gap_penalty: extending_gap_penalty,
    }
  }
}

#[inline]
pub fn durbin_algo(seq_pair: &SsPair, sa_sps: &SaScoringParams) -> ProbMat {
  let seq_len_pair = (seq_pair.0.len(), seq_pair.1.len());
  let (sa_part_func_mats, scale_param_mat) = get_sa_part_func_mats_and_scale_param_mat(seq_pair, &seq_len_pair, sa_sps);
  get_char_align_prob_mat(&sa_part_func_mats, &seq_len_pair, &scale_param_mat)
}

#[inline]
pub fn get_cap_mat_and_unaligned_char_psp(sp: &SsPair, sa_sps: &SaScoringParams) -> (ProbMat, ProbSeqPair) {
  let slp = (sp.0.len(), sp.1.len());
  let cap_mat = durbin_algo(sp, sa_sps);
  let mut ucp_seq_pair = (vec![0.; slp.0], vec![0.; slp.1]);
  for i in 0 .. slp.0 {
    let mut ucp = 1.;
    for j in 0 .. slp.1 {
      ucp -= cap_mat[i][j];
    }
    ucp_seq_pair.0[i] = ucp;
  }
  for j in 0 .. slp.1 {
    let mut ucp = 1.;
    for i in 0 .. slp.0 {
      ucp -= cap_mat[i][j];
    }
    ucp_seq_pair.1[j] = ucp;
  }
  (cap_mat, ucp_seq_pair)
}

#[inline]
pub fn get_sa_part_func_mats_and_scale_param_mat(sp: &SsPair, slp: &(usize, usize), sa_sps: &SaScoringParams) -> (SaPartFuncMats, ScaleParamMat) {
  let mut exp_sa_sps = sa_sps.clone();
  for ca_score in exp_sa_sps.ca_sm.values_mut() {
    *ca_score = ca_score.exp();
  };
  exp_sa_sps.opening_gap_penalty = exp_sa_sps.opening_gap_penalty.exp();
  exp_sa_sps.extending_gap_penalty = exp_sa_sps.extending_gap_penalty.exp();
  let mut sa_part_func_mats = SaPartFuncMats::new(slp);
  let mut scale_param_mat = vec![vec![0.; slp.1 + 2]; slp.0 + 2];
  scale_param_mat[0][0] = 1.;
  sa_part_func_mats.sa_forward_part_func_mat_4_char_align[0][0] = 1.;
  for i in 1 .. slp.0 + 1 {
    scale_param_mat[i][0] = exp_sa_sps.opening_gap_penalty * get_egp(&(1, i - 1), sa_sps).exp();
    sa_part_func_mats.sa_forward_part_func_mat_4_gap_1[i][0] = 1.;
  }
  for i in 1 .. slp.1 + 1 {
    scale_param_mat[0][i] = exp_sa_sps.opening_gap_penalty * get_egp(&(1, i - 1), sa_sps).exp();
    sa_part_func_mats.sa_forward_part_func_mat_4_gap_2[0][i] = 1.;
  }
  for i in 1 .. slp.0 + 2 {
    for j in 1 .. slp.1 + 2 {
      let ca_score = if i == slp.0 + 1 && j == slp.1 + 1 {
        1.
      } else if i == slp.0 + 1 || j == slp.1 + 1 {
        0.
      } else {
        exp_sa_sps.ca_sm[&(sp.0[i - 1], sp.1[j - 1])]
      };
      let mut scale_param = 0.;
      let forward_part_func = (sa_part_func_mats.sa_forward_part_func_mat_4_char_align[i - 1][j - 1] + sa_part_func_mats.sa_forward_part_func_mat_4_gap_1[i - 1][j - 1] + sa_part_func_mats.sa_forward_part_func_mat_4_gap_2[i - 1][j - 1]) * ca_score;
      scale_param += forward_part_func;
      sa_part_func_mats.sa_forward_part_func_mat_4_char_align[i][j] = forward_part_func;
      if i < slp.0 + 1 && j < slp.1 + 1 {
        let forward_part_func = sa_part_func_mats.sa_forward_part_func_mat_4_char_align[i - 1][j] * exp_sa_sps.opening_gap_penalty + sa_part_func_mats.sa_forward_part_func_mat_4_gap_1[i - 1][j] * exp_sa_sps.extending_gap_penalty;
        scale_param += forward_part_func;
        sa_part_func_mats.sa_forward_part_func_mat_4_gap_1[i][j] = forward_part_func;
        let forward_part_func = sa_part_func_mats.sa_forward_part_func_mat_4_char_align[i][j - 1] * exp_sa_sps.opening_gap_penalty + sa_part_func_mats.sa_forward_part_func_mat_4_gap_2[i][j - 1] * exp_sa_sps.extending_gap_penalty;
        scale_param += forward_part_func;
        sa_part_func_mats.sa_forward_part_func_mat_4_gap_2[i][j] = forward_part_func;
      }
      scale_param_mat[i][j] = scale_param;
      sa_part_func_mats.sa_forward_part_func_mat_4_char_align[i][j] /= scale_param;
      if i < slp.0 + 1 && j < slp.1 + 1 {
        sa_part_func_mats.sa_forward_part_func_mat_4_gap_1[i][j] /= scale_param;
        sa_part_func_mats.sa_forward_part_func_mat_4_gap_2[i][j] /= scale_param;
      }
    }
  }
  sa_part_func_mats.sa_backward_part_func_mat_4_char_align[slp.0 + 1][slp.1 + 1] = 1. / scale_param_mat[slp.0 + 1][slp.1 + 1];
  for i in 0 .. slp.0 {
    sa_part_func_mats.sa_forward_part_func_mat_4_gap_1[i][slp.1 + 1] = exp_sa_sps.opening_gap_penalty * get_egp(&(i, slp.0 - 1), sa_sps).exp() / scale_param_mat[i][slp.1 + 1];
  }
  for i in 0 .. slp.1 {
    sa_part_func_mats.sa_forward_part_func_mat_4_gap_1[slp.0 + 1][i] = exp_sa_sps.opening_gap_penalty * get_egp(&(i, slp.1 - 1), sa_sps).exp() / scale_param_mat[slp.0 + 1][i];
  }
  for i in (0 .. slp.0 + 1).rev() {
    for j in (0 .. slp.1 + 1).rev() {
      let ca_score = if i == slp.0 && j == slp.1 {
        1.
      } else if i == slp.0 || j == slp.1 {
        0.
      } else {
        exp_sa_sps.ca_sm[&(sp.0[i], sp.1[j])]
      };
      let scale_param = scale_param_mat[i][j];
      sa_part_func_mats.sa_backward_part_func_mat_4_char_align[i][j] = (sa_part_func_mats.sa_backward_part_func_mat_4_char_align[i + 1][j + 1] * ca_score + (sa_part_func_mats.sa_backward_part_func_mat_4_gap_1[i + 1][j] + sa_part_func_mats.sa_backward_part_func_mat_4_gap_2[i][j + 1]) * exp_sa_sps.opening_gap_penalty) / scale_param;
      if i > 0 && j > 0 {
        sa_part_func_mats.sa_backward_part_func_mat_4_gap_1[i][j] = (sa_part_func_mats.sa_backward_part_func_mat_4_char_align[i + 1][j + 1] * ca_score + sa_part_func_mats.sa_backward_part_func_mat_4_gap_1[i + 1][j] * exp_sa_sps.extending_gap_penalty) / scale_param;
        sa_part_func_mats.sa_backward_part_func_mat_4_gap_2[i][j] = (sa_part_func_mats.sa_backward_part_func_mat_4_char_align[i + 1][j + 1] * ca_score + sa_part_func_mats.sa_backward_part_func_mat_4_gap_2[i][j + 1] * exp_sa_sps.extending_gap_penalty) / scale_param;
      }
    }
  }
  (sa_part_func_mats, scale_param_mat)
}

#[inline]
fn get_char_align_prob_mat(sa_part_func_mats: &SaPartFuncMats, slp: &(usize, usize), scale_param_mat: &ScaleParamMat) -> ProbMat {
  let mut cap_mat = vec![vec![0.; slp.1]; slp.0];
  for i in 1 .. slp.0 + 1 {
    for j in 1 .. slp.1 + 1 {
      cap_mat[i - 1][j - 1] = scale_param_mat[i][j] * sa_part_func_mats.sa_forward_part_func_mat_4_char_align[i][j] * sa_part_func_mats.sa_backward_part_func_mat_4_char_align[i][j];
    }
  }
  cap_mat
}

#[inline]
fn get_egp(pp: &PosPair, sa_sps: &SaScoringParams) -> SaScore {
  (pp.1 + 1 - pp.0) as SaScore * sa_sps.extending_gap_penalty
}