lefse.py

import os,sys,math,pickle
import random as lrand
import rpy2.robjects as robjects
import argparse
import numpy
#import svmutil

def init():
    lrand.seed(1982)
    robjects.r('library(splines)')
    robjects.r('library(stats4)')
    robjects.r('library(survival)')
    robjects.r('library(mvtnorm)')
    robjects.r('library(modeltools)')
    robjects.r('library(coin)')
    robjects.r('library(MASS)')

def get_class_means(class_sl,feats):
    means = {}
    clk = list(class_sl.keys())
    for fk,f in feats.items():
        means[fk] = [numpy.mean((f[class_sl[k][0]:class_sl[k][1]])) for k in clk]
    return clk,means

def save_res(res,filename):
    with open(filename, 'w') as out:
        for k,v in res['cls_means'].items():
            out.write(k+"\t"+str(math.log(max(max(v),1.0),10.0))+"\t")
            if k in res['lda_res_th']:
                for i,vv in enumerate(v):
                    if vv == max(v):
                        out.write(str(res['cls_means_kord'][i])+"\t")
                        break
                out.write(str(res['lda_res'][k]))
            else: out.write("\t")
            out.write( "\t" + (res['wilcox_res'][k] if 'wilcox_res' in res and k in res['wilcox_res'] else "-")+"\n")

def load_data(input_file, nnorm = False):
    with open(input_file, 'rb') as inputf:
        inp = pickle.load(inputf)
    if nnorm: return inp['feats'],inp['cls'],inp['class_sl'],inp['subclass_sl'],inp['class_hierarchy'],inp['norm']
    else: return inp['feats'],inp['cls'],inp['class_sl'],inp['subclass_sl'],inp['class_hierarchy']

def load_res(input_file):
    with open(input_file, 'rb') as inputf:
        inp = pickle.load(inputf)
    return inp['res'],inp['params'],inp['class_sl'],inp['subclass_sl']


def test_kw_r(cls,feats,p,factors):
    robjects.globalenv["y"] = robjects.FloatVector(feats)
    for i,f in enumerate(factors):
        robjects.globalenv['x'+str(i+1)] = robjects.FactorVector(robjects.StrVector(cls[f]))
    fo = "y~x1"
    #for i,f in enumerate(factors[1:]):
    #   if f == "subclass" and len(set(cls[f])) <= len(set(cls["class"])): continue
    #   if len(set(cls[f])) == len(cls[f]): continue
    #   fo += "+x"+str(i+2)
    kw_res = robjects.r('kruskal.test('+fo+',)$p.value')
    return float(tuple(kw_res)[0]) < p, float(tuple(kw_res)[0])

def test_rep_wilcoxon_r(sl,cl_hie,feats,th,multiclass_strat,mul_cor,fn,min_c,comp_only_same_subcl,curv=False):
    comp_all_sub = not comp_only_same_subcl
    tot_ok =  0
    alpha_mtc = th
    all_diff = []
    for pair in [(x,y) for x in cl_hie.keys() for y in cl_hie.keys() if x < y]:
        dir_cmp = "not_set" #
        l_subcl1, l_subcl2 = (len(cl_hie[pair[0]]), len(cl_hie[pair[1]]))
        if mul_cor != 0: alpha_mtc = th*l_subcl1*l_subcl2 if mul_cor == 2 else 1.0-math.pow(1.0-th,l_subcl1*l_subcl2)
        ok = 0
        curv_sign = 0
        first = True
        for i,k1 in enumerate(cl_hie[pair[0]]):
            br = False
            for j,k2 in enumerate(cl_hie[pair[1]]):
                if not comp_all_sub and k1[len(pair[0]):] != k2[len(pair[1]):]:
                    ok += 1
                    continue
                cl1 = feats[sl[k1][0]:sl[k1][1]]
                cl2 = feats[sl[k2][0]:sl[k2][1]]
                med_comp = False
                if len(cl1) < min_c or len(cl2) < min_c:
                    med_comp = True
                sx,sy = numpy.median(cl1),numpy.median(cl2)
                if cl1[0] == cl2[0] and len(set(cl1)) == 1 and  len(set(cl2)) == 1:
                    tres, first = False, False
                elif not med_comp:
                    robjects.globalenv["x"] = robjects.FloatVector(cl1+cl2)
                    robjects.globalenv["y"] = robjects.FactorVector(robjects.StrVector(["a" for a in cl1]+["b" for b in cl2]))
                    pv = float(robjects.r('pvalue(wilcox_test(x~y,data=data.frame(x,y)))')[0])
                    tres = pv < alpha_mtc*2.0
                if first:
                    first = False
                    if not curv and ( med_comp or tres ):
                        dir_cmp = sx < sy
                        #if sx == sy: br = True
                    elif curv:
                        dir_cmp = None
                        if med_comp or tres:
                            curv_sign += 1
                            dir_cmp = sx < sy
                    else: br = True
                elif not curv and med_comp:
                    if ((sx < sy) != dir_cmp or sx == sy): br = True
                elif curv:
                    if tres and dir_cmp == None:
                        curv_sign += 1
                        dir_cmp = sx < sy
                    if tres and dir_cmp != (sx < sy):
                        br = True
                        curv_sign = -1
                elif not tres or (sx < sy) != dir_cmp or sx == sy: br = True
                if br: break
                ok += 1
            if br: break
        if curv: diff = curv_sign > 0
        else: diff = (ok == len(cl_hie[pair[1]])*len(cl_hie[pair[0]])) # or (not comp_all_sub and dir_cmp != "not_set")
        if diff: tot_ok += 1
        if not diff and multiclass_strat: return False
        if diff and not multiclass_strat: all_diff.append(pair)
    if not multiclass_strat:
        tot_k = len(list(cl_hie.keys()))
        for k in cl_hie.keys():
            nk = 0
            for a in all_diff:
                if k in a: nk += 1
            if nk == tot_k-1: return True
        return False
    return True



def contast_within_classes_or_few_per_class(feats,inds,min_cl,ncl):
    ff = list(zip(*[v for n,v in feats.items() if n != 'class']))
    cols = [ff[i] for i in inds]
    cls = [feats['class'][i] for i in inds]
    if len(set(cls)) < ncl:
        return True
    for c in set(cls):
        if cls.count(c) < min_cl:
            return True
        cols_cl = [x for i,x in enumerate(cols) if cls[i] == c]
        for i,col in enumerate(zip(*cols_cl)):
            if (len(set(col)) <= min_cl and min_cl > 1) or (min_cl == 1 and len(set(col)) <= 1):
                return True
    return False

def test_lda_r(cls,feats,cl_sl,boots,fract_sample,lda_th,tol_min,nlogs):
    fk = list(feats.keys())
    means = dict([(k,[]) for k in feats.keys()])
    feats['class'] = list(cls['class'])
    clss = list(set(feats['class']))

    for uu,k in enumerate(fk):
        if k == 'class':
            continue

        ff = [(feats['class'][i],v) for i,v in enumerate(feats[k])]

        for c in clss:
            if len(set([float(v[1]) for v in ff if v[0] == c])) > max(float(feats['class'].count(c))*0.5,4):
                continue

            for i,v in enumerate(feats[k]):
                if feats['class'][i] == c:
                    feats[k][i] = math.fabs(feats[k][i] + lrand.normalvariate(0.0,max(feats[k][i]*0.05,0.01)))

    rdict = {}

    for a,b in feats.items():
        if a == 'class' or a == 'subclass' or a == 'subject':
            rdict[a] = robjects.StrVector(b)
        else:
            rdict[a] = robjects.FloatVector(b)

    robjects.globalenv["d"] = robjects.DataFrame(rdict)
    lfk = len(feats[fk[0]])
    rfk = int(float(len(feats[fk[0]]))*fract_sample)
    f = "class ~ "+fk[0]

    for k in fk[1:]:
        f += " + " + k.strip()

    ncl = len(set(cls['class']))
    min_cl = int(float(min([cls['class'].count(c) for c in set(cls['class'])]))*fract_sample*fract_sample*0.5)
    min_cl = max(min_cl,1)
    pairs = [(a,b) for a in set(cls['class']) for b in set(cls['class']) if a > b]

    for k in fk:
        for i in range(boots):
            means[k].append([])

    for i in range(boots):
        for rtmp in range(1000):
            rand_s = [lrand.randint(0,lfk-1) for v in range(rfk)]
            if not contast_within_classes_or_few_per_class(feats,rand_s,min_cl,ncl):
                break

        rand_s = [r+1 for r in rand_s]
        means[k][i] = []

        for p in pairs:
            robjects.globalenv["rand_s"] = robjects.IntVector(rand_s)
            robjects.globalenv["sub_d"] = robjects.r('d[rand_s,]')
            z = robjects.r('z <- suppressWarnings(lda(as.formula('+f+'),data=sub_d,tol='+str(tol_min)+'))')
            robjects.r('w <- z$scaling[,1]')
            robjects.r('w.unit <- w/sqrt(sum(w^2))')
            robjects.r('ss <- sub_d[,-match("class",colnames(sub_d))]')

            if 'subclass' in feats:
                robjects.r('ss <- ss[,-match("subclass",colnames(ss))]')

            if 'subject' in feats:
                robjects.r('ss <- ss[,-match("subject",colnames(ss))]')

            robjects.r('xy.matrix <- as.matrix(ss)')
            robjects.r('LD <- xy.matrix%*%w.unit')
            robjects.r('effect.size <- abs(mean(LD[sub_d[,"class"]=="'+p[0]+'"]) - mean(LD[sub_d[,"class"]=="'+p[1]+'"]))')
            scal = robjects.r('wfinal <- w.unit * effect.size')
            rres = robjects.r('mm <- z$means')
            rowns = list(rres.rownames)
            lenc = len(list(rres.colnames))
            coeff = [abs(float(v)) if not math.isnan(float(v)) else 0.0 for v in scal]
            res = dict([(pp,[float(ff) for ff in rres.rx(pp,True)] if pp in rowns else [0.0]*lenc ) for pp in [p[0],p[1]]])

            for j,k in enumerate(fk):
                gm = abs(res[p[0]][j] - res[p[1]][j])
                means[k][i].append((gm+coeff[j])*0.5)

    res = {}

    for k in fk:
        m = max([numpy.mean([means[k][kk][p] for kk in range(boots)]) for p in range(len(pairs))])
        res[k] = math.copysign(1.0,m)*math.log(1.0+math.fabs(m),10)

    return res,dict([(k,x) for k,x in res.items() if math.fabs(x) > lda_th])


def test_svm(cls,feats,cl_sl,boots,fract_sample,lda_th,tol_min,nsvm):
    return None
"""
    fk = feats.keys()
    clss = list(set(cls['class']))
    y = [clss.index(c)*2-1 for c in list(cls['class'])]
    xx = [feats[f] for f in fk]
    if nsvm:
        maxs = [max(v) for v in xx]
        mins = [min(v) for v in xx]
        x = [ dict([(i+1,(v-mins[i])/(maxs[i]-mins[i])) for i,v in enumerate(f)]) for f in zip(*xx)]
    else: x = [ dict([(i+1,v) for i,v in enumerate(f)]) for f in zip(*xx)]

    lfk = len(feats[fk[0]])
    rfk = int(float(len(feats[fk[0]]))*fract_sample)
    mm = []

    best_c = svmutil.svm_ms(y, x, [pow(2.0,i) for i in range(-5,10)],'-t 0 -q')
    for i in range(boots):
        rand_s = [lrand.randint(0,lfk-1) for v in range(rfk)]
        r = svmutil.svm_w([y[yi] for yi in rand_s], [x[xi] for xi in rand_s], best_c,'-t 0 -q')
        mm.append(r[:len(fk)])
    m = [numpy.mean(v) for v in zip(*mm)]
    res = dict([(v,m[i]) for i,v in enumerate(fk)])
    return res,dict([(k,x) for k,x in res.items() if math.fabs(x) > lda_th])
"""