aboutsummaryrefslogtreecommitdiff
path: root/bin/cluster.hs
blob: a0032278fb2e8dab9d05bc6a768783b1974d29ce (plain)
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
{-# LANGUAGE ViewPatterns #-}
{-# LANGUAGE ScopedTypeVariables #-}
{-# LANGUAGE LambdaCase #-}

module Main where

import Control.Monad
import Data.Fixed (mod')
import Data.Foldable (toList)
import Data.List
import qualified Data.Map as M
import GHC.Compact
import Numeric.Log hiding (sum)
import Options.Applicative
import PPL hiding (binom)
import System.Random (mkStdGen, setStdGen, StdGen, random, split)
import qualified Streaming as S
import qualified Streaming.Prelude as S
import Streaming.Prelude (each, fold, Of, Stream, yield)
import System.IO ( hSetBuffering, BufferMode(NoBuffering), stdout)
import System.ProgressBar

cumsum = scanl1 (+)

first f xs = snd . head . filter (f . fst) $ zip xs [0 ..]

stirling 0 = 1
stirling 1 = 1
stirling n = Exp $ n * (log n - 1) + log (sqrt (2 * pi)) + log n / 2

logFromInt = logFrom . fromIntegral

logFrom = Exp . log

binom :: Int -> Int -> Double -> Log Double
binom n@(logFromInt -> n') k@(logFromInt -> k') p =
  n `choose` k * logFrom p ** k' * logFrom (1 - p) ** (n' - k')
  where
    choose (fromIntegral -> n) (fromIntegral -> k) = stirling n / stirling k / stirling (n - k)

-- (infinite) binary trees
data Tree a = Tree a (Tree a) (Tree a)
  deriving (Show)

instance Foldable Tree where
  foldMap f t = bftrav [t]
    where
      bftrav [] = mempty
      bftrav ((Tree a l r) : ts) = f a <> bftrav (ts <> [l, r])

{-# INLINE pairs #-}
pairs (a : b : rs) = (a, b) : pairs rs
pairs [] = []
pairs _ = error "unexpected number of columns, expecting pairs"

-- Infinite trees from infinite lists
-- NB: it's harder to partition a list so that it folds back to
-- equivalence. It doesn't really matter here since we're only
-- unfolding random uniforms anyway.
treeFromList (x : xs) = Tree x (treeFromList lpart) (treeFromList rpart)
  where
    (lpart, rpart) = unzip $ pairs xs

-- Constrain trees so leaves sum to node value
normTree :: Tree Double -> Tree Double
normTree (Tree x (Tree u l r) (Tree v l' r')) =
  let s = x / (u + v)
   in Tree x (normTree $ Tree (s * u) l r) (normTree $ Tree (s * v) l' r')

drawTreeProbs = toList . normTree . treeFromList <$> iid uniform

model :: [[Int]] -> Meas ([[Double]], [Int])
model xs = do
  (ps, params, clusters) <- sample $ do
    -- Sample hyperparameters
    a <- bounded 1 10

    -- CRP style
    dir <- cumsum <$> dirichletProcess a
    rs <- iid uniform
    ps <- iid drawTreeProbs
    let clusters = map (\r -> first (>= r) dir) rs
        params = map (transpose ps !!) clusters
    pure (ps, params, clusters)
  mapM_ scoreLog $ zipWith likelihood params xs
  let cls = take (length xs) clusters
      k = maximum cls + 1
      n = length (head xs) `div` 2
  pure (map (take k) $ take n ps, cls)
  where
    likelihood ps cnts = product $ zipWith go ps (pairs cnts)
      where
        go p (c, d) = binom d c p

-- Tabulate list
tabulate xs = M.elems $ M.fromListWith (+) [(c, 1) | c <- xs]

-- Command line args
data Options = Options
  { seed :: Int,
    nsamples :: Int,
    mhfrac :: Double,
    input :: FilePath,
    propsPath :: FilePath,
    clusterPath :: FilePath
  }

main = run =<< execParser opts
  where
    opts =
      info
        (parser <**> helper)
        (fullDesc <> progDesc "Infer a phylogeny from SNV calls in multiple samples" <> header "phylogey - Bayesian phylogeny inference")
    parser =
      Options <$> option auto (long "seed" <> short 's' <> help "random seed" <> showDefault <> value 42 <> metavar "INT")
        <*> option auto (long "nsamples" <> short 'n' <> help "number of samples from posterior" <> value 100000 <> metavar "INT" <> showDefault)
        <*> option probability (long "mhfrac" <> short 'm' <> help "Metropolis-Hastings mutation probability" <> value 0.3 <> metavar "(0,1]" <> showDefault)
        <*> argument str (metavar "INPUT")
        <*> argument str (metavar "PROPS")
        <*> argument str (metavar "TREE")

    probability = eitherReader $ \arg -> case reads arg of
      [(r, "")] -> if r <= 1 && r > 0 then Right r else Left "mhfrac not a valid probability"
      _ -> Left "mhfrac not a valid probability"


takeWithProgress :: S.MonadIO m => Int -> Stream (Of a) m r -> Stream (Of a) m ()
takeWithProgress n str = do
  pb <- S.liftIO $ newProgressBar defStyle 10 (Progress 0 n ())
  S.mapM (update pb) $ S.take n str
  where
    update pb x = do
      S.liftIO $ incProgress pb 1
      pure x

run opts = do
  (hdr : lines) <- lines <$> readFile (input opts)
  let dbl = round . read :: String -> Int
      g = mkStdGen $ seed opts
  parsed <- compact $ map (map dbl . tail . words) lines
  hSetBuffering stdout NoBuffering
  ((ps, cl), _) <- S.fold_ (\l r -> if mml l < mml r then l else r) (([[]],[]), -1/0) id . takeWithProgress (nsamples opts) $ mh g (mhfrac opts) (model $ getCompact parsed)
  writeFile (propsPath opts) . unlines $ map (intercalate "," . map show) ps
  writeFile (clusterPath opts) . unlines $ map show cl
  where
    mml ((ps, cl), lik) = sum' (sum' (log . (+ 1))) ps + sum' (log . (+ 1)) tab - sum' (ln . stirling) tab - ln lik
      where
        tab = tabulate cl
        sum' f = sum . map f