frontends4.h 87.7 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
/* Spell-QTL  Software suite for the QTL analysis of modern datasets.
 * Copyright (C) 2016,2017  Damien Leroux <damien.leroux@inra.fr>, Sylvain Jasson <sylvain.jasson@inra.fr>
 *
 * This program is free software: you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation, either version 3 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program.  If not, see <http://www.gnu.org/licenses/>.
 */

18
19
20
21
22
23
24
#ifndef _SPEL_FRONTENDS_H_
#define _SPEL_FRONTENDS_H_

#include "cache2.h"
#include "basic_data.h"
#include "model.h"
/*#include "model/tests.h"*/
25
#include <regex>
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
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245

#include <boost/math/distributions/normal.hpp> // for normal_distribution
  using boost::math::normal; // typedef provides default type is double.
  using boost::math::cdf;
  using boost::math::mean;
  using boost::math::variance;
  using boost::math::quantile;
  using boost::math::complement;


typedef std::pair<const chromosome*, double> selected_locus;

inline bool operator < (const selected_locus& sl1, const selected_locus& sl2) { return sl1.first < sl2.first || (sl1.first == sl2.first && sl1.second < sl2.second); }

inline std::ostream& operator << (std::ostream& os, const selected_locus& sl) { return os << sl.first->name << ':' << sl.second; }


struct chromosome_search_domain {
    const chromosome* chrom;
    std::vector<double> loci;

    chromosome_search_domain(const chromosome* c, const std::vector<double>& l) : chrom(c), loci(l) {}

    struct const_iterator {
        const chromosome_search_domain* this_csd;
        std::vector<double>::const_iterator i;
        const_iterator() : this_csd(NULL), i(__()) {}
        const_iterator(const chromosome_search_domain* t, const std::vector<double>::const_iterator& i_) : this_csd(t), i(i_) {}
        bool operator == (const const_iterator& other) const { return i == other.i; }
        bool operator != (const const_iterator& other) const { return i != other.i; }
        /*bool operator < (const const_iterator& other) const { return i < other.i; }*/
        /*size_t operator - (const const_iterator& other) const { return i - other.i; }*/
        std::pair<const chromosome*, double> operator * () const { return {this_csd->chrom, *i}; }
        const_iterator& operator ++ () { ++i; return *this; }
        const_iterator& operator -- () { --i; return *this; }
        static std::vector<double>::const_iterator __() { static std::vector<double> _; return _.end(); }
    };

    const_iterator begin() const { return {this, loci.begin()}; }
    const_iterator end() const { return {this, loci.end()}; }
    const_iterator cbegin() const { return {this, loci.begin()}; }
    const_iterator cend() const { return {this, loci.end()}; }
};


typedef std::vector<chromosome_search_domain> genome_search_domain;

struct gsd_iterator {
    std::vector<chromosome_search_domain>::const_iterator csd_i, csd_j;
    chromosome_search_domain::const_iterator i, j;

    bool operator == (const gsd_iterator& other) const { return csd_i == other.csd_i && i == other.i; }
    bool operator != (const gsd_iterator& other) const { return csd_i != other.csd_i || i != other.i; }

    gsd_iterator&
        operator ++ ()
        {
            ++i;
            if (i == j) {
                MSG_DEBUG("at end of chromosome!");
                ++csd_i;
                if (csd_i != csd_j) {
                    i = csd_i->begin();
                    j = csd_i->end();
                } else {
                    i = {};
                    j = {};
                }
            }
            return *this;
        }

    std::pair<const chromosome*, double> operator * () const { return *i; }
};

namespace std {
    inline gsd_iterator begin(const genome_search_domain& gsd) { return {gsd.begin(), gsd.end(), gsd.begin()->begin(), gsd.begin()->end()}; }
    inline gsd_iterator end(const genome_search_domain& gsd) { return {gsd.end(), gsd.end(), {}, {}}; }
}


typedef std::vector<selected_locus> locus_set;

typedef std::vector<locus_set> model_descriptor;

std::pair<bool, double>
detect_strongest_qtl(chromosome_value chr, const locus_key& lk,
                     const model& M0, const std::vector<double> pos);

MatrixXd
ftest_along_chromosome(chromosome_value chr, const locus_key& lk,
                       const model& M0, const std::vector<double> pos);


/* Definitions:
 * - cofactor: isolated POP (single chromosome, single locus)
 * - QTLs: joint POP (single chromosome, single or multiple loc(i)us)
 *
 *
 * Configurations:
 * - with/without Dominance                          D
 * - with/without Constraints (can/can't estimate)   WC
 * - Joint/Single POP computation mode               JS
 *
 *
 * Steps:                           D   JS  WC
 * - establish skeleton                      
 *   - manual (marker list)                  
 *   - by step                               
 * - discover cofactors                 S    
 *   - manual                           S    
 *   - forward                          S    
 *   - backward                         S    
 * - detect QTLs                    ?   J   Y
 *   - CIM-                         ?   J   Y
 *   - iQTLm                        ?   J   Y
 *   - iQTLm++                      ?   J   Y
 * - OPTIONALLY analyze epistasis   ?   J   Y
 * - estimate parameters            
 *
 *
 * Operations:
 * - select chromosome
 * - cofactors to QTLs for the current chromosome
 * - QTLs to cofactors for the current chromosome
 * - test along the chromosome
 * - test along all chromosomes
 * - add cofactor (if current chromosome in product probability mode)
 * - add QTL (if current chromosome in joint probability mode)
 * - remove cofactor/QTL
 */


struct signal_display {
#ifdef SIGNAL_DISPLAY_ONELINER
    static const char* tick(double x)
    {
        static const char* ticks[9] = { " ", "\u2581", "\u2582", "\u2583", "\u2584", "\u2585", "\u2586", "\u2587", "\u2588" };
        return ticks[x < 0. ? 0
                            : x >= 1. ? 8
                                      : ((int) floor(x * 9))];
    }

    VectorXd values;
    int imax_;
    bool above_;

    signal_display(const VectorXd& v, int imax, bool above)
        : values(v.innerSize()), imax_(imax), above_(above)
    {
        values = v;
#if 0
        int sig_cols = msg_handler_t::termcols() - 3;
        MSG_DEBUG("values.innerSize = " << values.innerSize());
        MSG_QUEUE_FLUSH();
        while (values.innerSize() >= sig_cols) {
            if (values.innerSize() & 1) {
                int sz = values.innerSize();
                values.conservativeResize(sz + 1);
                values(sz) = values(sz - 1);
            }
            int i = values.innerSize() >> 1;
            values = values.transpose() * kroneckerProduct(MatrixXd::Identity(i, i), MatrixXd::Constant(1, 2, .5));
            MSG_DEBUG("values.innerSize = " << values.innerSize());
            MSG_QUEUE_FLUSH();
        }
#endif
        double vmin = values.minCoeff();
        double vmax = values.maxCoeff();
        if (vmin == vmax) {
            values = (values.array() - vmin).matrix();
        } else {
            values = ((values.array() - vmin) / (vmax - vmin)).matrix();
        }
    }

    friend std::ostream& operator << (std::ostream& os, const signal_display& sd)
    {
        os << _WHITE << '[';
        for (int i = 0; i < sd.values.innerSize(); ++i) {
            if (i == sd.imax_) {
                os << (sd.above_ ? _GREEN : _RED);
            }
            os << tick(sd.values(i));
            if (i == sd.imax_) {
                os << _WHITE;
            }
        }
        return os << ']' << _NORMAL;
    }
#else
    braille_grid grid;

    signal_display(const chromosome& chr, const std::vector<double>& X, const VectorXd& y, int imax, double threshold)
        : grid(build(chr, X, y, imax, threshold))
    {}

    braille_grid
        build(const chromosome& chr, const std::vector<double>& X, const VectorXd& y, int imax, double threshold)
        {
            std::vector<double> Y(y.data(), y.data() + y.size());
            int padding_left = 0;
            int W = (int) (msg_handler_t::termcols() * .8);
            if (W > 1000) {
                W = 80;
            }
            braille_grid chr_map = chr.pretty_print(W, {}, {}, padding_left, false);

            braille_plot plot(W - padding_left, 5, 0, X.back(), 0, std::max(threshold, y(imax)));
            plot.plot(X, Y);
            plot.hline(threshold, 1, 1, 0, 255, 0);
            bool above = y(imax) > threshold;
            plot.vline(X[imax], 1, 0, above ? 0 : 255, above ? 255 : 0, 0);
            return plot.compose_vert(true, chr_map, false);
        }

    friend
        std::ostream&
        operator << (std::ostream& os, const signal_display& sd)
        {
246
247
248
            std::stringstream tmp;
            tmp << sd.grid;
            return os << tmp.str();
249
250
251
252
253
        }
#endif
};


254
255
struct model_manager;

256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
struct QTL {
    std::string chromosome;
    double locus;
    std::vector<double> LOD_loci;
    std::vector<double> LOD;

    QTL(const std::string& n, double l, const std::vector<double>& x, const MatrixXd& y)
        : chromosome(n), locus(l), LOD_loci(x), LOD(y.data(), y.data() + y.size())
    {
        /*MSG_DEBUG("QTL at " << chromosome << ':' << locus);*/
        /*MSG_DEBUG(y);*/
        /*MSG_DEBUG(MATRIX_SIZE(y));*/
        /*MSG_DEBUG("" << LOD);*/
    }

    static
        double
        interpolate(double x0, double y0, double x1, double y1, double yT)
        {
            double delta_x = x1 - x0;
            double delta_y = y1 - y0;
            return delta_x * (yT - y0) / delta_y + x0;
        }

    std::pair<double, double>
281
        confidence_interval(const std::string &trait, const std::vector<QTL> &selection, ComputationType lod_test_type);
282
283
284

    std::pair<double, double>
        confidence_interval();
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
#if 0
        {
            /*MSG_DEBUG_INDENT_EXPR("[Confidence interval] ");*/
            /*MSG_DEBUG("LOD: " << LOD);*/
            double maxLOD = *std::max_element(LOD.begin(), LOD.end());
            double lod_cap = maxLOD - 1.5;
            /*MSG_DEBUG("max=" << maxLOD << " threshold=" << lod_cap);*/
            int i;
            for (i = 0; i < (int) LOD_loci.size() && LOD_loci[i] < locus && LOD[i] < lod_cap; ++i);
            /*MSG_DEBUG("LEFT i=" << i);*/
            double left;
            if (i > 0) {
                left = interpolate(LOD_loci[i - 1], LOD[i - 1], LOD_loci[i], LOD[i], lod_cap);
            } else {
                left = LOD_loci[i];
            }
            for (i = LOD_loci.size() - 1; i >= 0 && LOD_loci[i] > locus && LOD[i] < lod_cap; --i);
            /*MSG_DEBUG("RIGHT i=" << i);*/
            double right;
            if (i < (int) (LOD_loci.size() - 1)) {
                right = interpolate(LOD_loci[i], LOD[i], LOD_loci[i + 1], LOD[i + 1], lod_cap);
            } else {
                right = LOD_loci[i];
            }
            /*MSG_INFO("Confidence interval for " << chromosome << ':' << locus << " {" << left << ':' << right << '}');*/
            /*MSG_DEBUG_DEDENT;*/
            return {left, right};
        }
#endif
};


317
enum class AR: int { RSS=1, Rank=2, Test=4, Model=8, All=0xFF };
318

319
320
321
inline
bool
operator & (AR a, AR b) { return !!(((int) a) & ((int) b)); }
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403


enum probability_mode { Joint, Single };


struct analysis_report {

    bool output_rss;
    bool output_rank;
    bool output_test;
    bool output_model;

    std::string report_path;

    std::string trait_name;

    std::string full_path;

    file report_file;

    std::map<std::string, std::map<double, std::string>> poi;
    std::map<std::string, std::map<double, std::pair<double, double>>> roi;

    analysis_report(const std::string& path, AR what)
        : output_rss(what & AR::RSS), output_rank(what & AR::Rank), output_test(what & AR::Test), output_model(what & AR::Model)
        , report_path(path)
        , trait_name()
        , full_path()
        , report_file()
        , poi(), roi()
    {
        ensure_directories_exist(report_path);
    }

    ~analysis_report()
    {
        report_file.close();
        report_file.open(MESSAGE(report_path << "/full_map.txt"), std::fstream::out);
        for (const auto& chr: active_settings->map) {
            report_file << chr.pretty_print(200, poi[chr.name], roi[chr.name]) << std::endl;
        }
    }

    void attach_model_manager(model_manager& mm);
    void detach_model_manager(model_manager& mm);

    void report_trait(const std::string& /*name*/, const MatrixXd& values)
    {
        static Eigen::IOFormat trait_format(Eigen::FullPrecision, Eigen::DontAlignCols, "\t", "\n", "", "", "", "");
        std::string filename = MESSAGE(full_path << '/' << "trait_values.txt");
        ofile of(filename);
        of << values.format(trait_format);
        of.close();
    }

    void report_lod(const QTL& qtl)
    {
        std::string filename = MESSAGE(full_path << '/' << qtl.chromosome << ':' << qtl.locus << "_LOD.txt");
        ofile of(filename);
        for (size_t i = 0; i < qtl.LOD.size(); ++i) {
            of << qtl.LOD_loci[i] << '\t' << qtl.LOD[i] << std::endl;
        }
        of.close();
    }

    void report_model(const model& Mcurrent)
    {
        if (output_model) {
            Mcurrent.output_X_to_file(full_path);
            Mcurrent.output_XtX_inv_to_file(full_path);
        }
    }

    /*enum ComputationType { NoTest=0, FTest=1, FTestLOD=2, R2=4, Chi2=8, Mahalanobis=16 };*/
    std::string
        computation_type_to_string(ComputationType ct)
        {
            std::stringstream ret;
            if (ct & FTest) { ret << "_FTest"; }
            if (ct & FTestLOD) { ret << "_FTestLOD"; }
            if (ct & R2) { ret << "_R2"; }
            if (ct & Chi2) { ret << "_Chi2"; }
404
            if (ct & Chi2LOD) { ret << "_Chi2LOD"; }
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
            if (ct & Mahalanobis) { ret << "_Mahalanobis"; }
            return ret.str();
        }

    void report_computation(const model& Mcurrent, const chromosome* chrom_under_study, const computation_along_chromosome& cac, ComputationType ct, ComputationResults /*cr*/, const std::vector<double>& testpos, probability_mode pmode=Single)
    {
        if (output_test | output_rss | output_rank) {
            /*MSG_DEBUG(MATRIX_SIZE(cac.ftest_pvalue));*/
            /*MSG_DEBUG(MATRIX_SIZE(cac.rss));*/
            std::string path = MESSAGE(full_path << '/' << chrom_under_study->name);
            ensure_directories_exist(path);
            std::string filename
                = MESSAGE(path << '/' << Mcurrent.keys()
                        << (output_test ? computation_type_to_string(ct) : "")
                        << (output_rss ? "_RSS" : "")
                        << (output_rank ? "_Rank" : "")
                        << (pmode == Joint ? "_Joint" : "")
                        << ".txt"
                        );
            ofile f(filename);
            if (output_test) { f << '\t' << "Test"; }
            if (output_rss) { for (int i = 0; i < cac.rss.innerSize(); ++i) { f << '\t' << "RSS"; } }
            if (output_rank) { f << '\t' << "Rank"; }
            f << std::endl;
            for (size_t i = 0; i < testpos.size(); ++i) {
                f << testpos[i];
                if (output_test) {
                    switch(ct) {
                        case ComputationType::FTest:
                            f << '\t' << cac.ftest_pvalue(0, i);
                            break;
                        case ComputationType::FTestLOD:
                            f << '\t' << cac.ftest_lod(0, i);
                            break;
                        case ComputationType::Chi2:
                            f << '\t' << cac.chi2(0, i);
                            break;
442
443
444
                        case ComputationType::Chi2LOD:
                            f << '\t' << cac.chi2_lod(0, i);
                            break;
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
                        default:
                            /*last_computation = NULL;*/
                            ;
                    };
                }
                if (output_rss) {
                    for (int j = 0; j < cac.rss.innerSize(); ++j) {
                        f << '\t' << cac.rss(j, i);
                    }
                }
                if (output_rank) {
                    f << '\t' << cac.rank(i);
                }
                f << std::endl;
            }
            f.close();
        }
    }

    void report_final_model(model_manager& mm);

466
    void report_qtls(std::vector<QTL> &qtls, ComputationType lod_test_type);
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
};


typedef std::pair<double, double> forbidden_interval_type;
typedef std::vector<forbidden_interval_type> forbidden_interval_vector_type;
typedef std::map<chromosome_value, forbidden_interval_vector_type> forbidden_interval_map_type;

inline bool operator < (const forbidden_interval_type& fi1, const forbidden_interval_type& fi2) { return fi1.first < fi2.first; }

struct search_interval_type;

struct test_result {
    search_interval_type* this_interval;
    const chromosome* chrom;
    double locus;
    double test_value;
    int index;
    bool over_threshold;
485
486
487
488
    model_block_key pop_block_key;
    value<model_block_type> pop_block;
    model_block_key dom_block_key;
    value<model_block_type> dom_block;
489
490

    test_result()
491
        : chrom(NULL), locus(0), test_value(0), index(0), over_threshold(false), pop_block_key(), pop_block(), dom_block_key(), dom_block()
492
493
    {}

494
495
    test_result(search_interval_type* ti, const chromosome* c, double l, double tv, int i, bool ot, const model_block_key& mbk, const value<model_block_type>& mb, const model_block_key& mbkd, const value<model_block_type>& mbd)
        : this_interval(ti), chrom(c), locus(l), test_value(tv), index(i), over_threshold(ot), pop_block_key(mbk), pop_block(mb), dom_block_key(mbkd), dom_block(mbd)
496
497
498
499
500
501
    {}

    test_result(const test_result& tr)
        : this_interval(tr.this_interval),
        chrom(tr.chrom), locus(tr.locus), test_value(tr.test_value),
        index(tr.index), over_threshold(tr.over_threshold),
502
503
        pop_block_key(tr.pop_block_key), pop_block(tr.pop_block),
        dom_block_key(tr.dom_block_key), dom_block(tr.dom_block)
504
505
506
507
508
509
510
511
512
513
514
    {}

    test_result&
        operator = (const test_result& tr)
        {
            this_interval = tr.this_interval;
            chrom = tr.chrom;
            locus = tr.locus;
            test_value = tr.test_value;
            index = tr.index;
            over_threshold = tr.over_threshold;
515
516
517
518
            pop_block_key = tr.pop_block_key;
            pop_block = tr.pop_block;
            dom_block_key = tr.dom_block_key;
            dom_block = tr.dom_block;
519
520
521
522
523
524
525
526
527
528
            return *this;
        }

    void reset()
    {
        chrom = NULL;
        locus = 0;
        test_value = 0;
        index = 0;
        over_threshold = false;
529
        /*block_key.selection.clear();*/
530
531
532
533
        pop_block_key.reset();
        pop_block = value<model_block_type>();
        dom_block_key.reset();
        dom_block = value<model_block_type>();
534
535
536
537
538
539
540
541
542
543
    }

    friend
        std::ostream& operator << (std::ostream& os, const test_result& tr)
        {
            os << "<result chrom=" << (tr.chrom ? tr.chrom->name : "nil")
                << " locus=" << tr.locus
                << " test=" << tr.test_value
                << " at=" << tr.index
                << " over?=" << tr.over_threshold
544
545
                << " block_key=" << tr.pop_block_key
                << " dominance_block_key=" << tr.dom_block_key
546
547
548
549
                << '>';
            return os;
        }

550
551
    bool
        operator < (const test_result& other) const
552
        {
553
            return test_value < other.test_value;
554
        }
555
556
557

    void
        select(model_manager& mm) const;
558
559
560
};


561
562
563
564
565
566
locus_probabilities_type
locus_probabilities(const context_key& ck, const locus_key& lk,
                    /*const MatrixXd& mgo,*/
                    int ind,
                    const std::vector<double>& loci);

567
568
struct search_interval_type {
    probability_mode mode;
569
    const chromosome* chrom;
570
571
572
573
574
575
576
577
578
579
580
    /* all positions in this interval */
    std::vector<double> all_positions;
    /* all USED positions in this interval */
    std::vector<double> positions;
    /* positions actually used in the segment test (all_positions \ selection \ forbidden_intervals) */
    std::vector<double> effective_positions;
    /* current selection (subset of base model's selection for this search interval) */
    locus_key selection;
    /* the model blocks along the chromosome currently under study */
    collection<model_block_type> locus_blocks;
    /* dominance matrices per locus per population */
581
    collection<parental_origin_per_locus_type> dominance_blocks;
582
583
    test_result local_max;

584
585
586
587
588
589
590
591
592
593
594
595
    search_interval_type(probability_mode pm, const chromosome* chr)
        : mode(pm)
        , chrom(chr)
        , all_positions()
        , positions()
        , effective_positions()
        , selection()
        , locus_blocks()
        , dominance_blocks()
        , local_max()
    {}

596
597
598
599
600
601
602
603
604
605
606
607
    search_interval_type(const search_interval_type& si)
        : mode(si.mode)
        , chrom(si.chrom)
        , all_positions(si.all_positions)
        , positions(si.positions)
        , effective_positions(si.effective_positions)
        , selection(si.selection)
        , locus_blocks(si.locus_blocks)
        , dominance_blocks(si.dominance_blocks)
        , local_max(si.local_max)
    {}

608
    void
609
        clear()
610
        {
611
612
613
            positions.clear();
            effective_positions.clear();
            locus_blocks.clear();
614
615
        }

616
617
618
    bool
        contains_position(double l) const
        {
619
            return l >= all_positions.front() && l <= all_positions.back()/* && std::find(all_positions.begin(), all_positions.end(), l) != all_positions.end()*/;
620
        }
621

622
623
624
625
626
627
628
629
630
    void
        compute_positions(forbidden_interval_vector_type::const_iterator& fi, forbidden_interval_vector_type::const_iterator& fj)
        {
            positions.clear();
            positions.reserve(all_positions.size());
            for (double d: all_positions) {
                for (; fi != fj && d > fi->second; ++fi);
                if (fi == fj || d < fi->first) {
                    positions.push_back(d);
631
632
633
                }
            }
        }
634
635

    void
636
637
638
        test(const collection<population_value>& all_pops, int i0, computation_along_chromosome& cac,
             value<ComputationType> vct, value<ComputationResults> vcr, size_t y_block_cols,
             const value<model>& Mcurrent, const value<model>& Mbase, double threshold)
639
640
641
642
643
        {
            locus_key lk;
            if (mode == Joint) {
                lk = selection;
            }
644
            /*MSG_DEBUG("Running test along interval with lk=" << lk);*/
645
            test(all_pops, i0, cac, vct, vcr, y_block_cols, Mcurrent, Mbase, lk, positions, threshold);
646
647
648
        }

    void
649
650
651
        test(const collection<population_value>& all_pops, int i0, computation_along_chromosome& cac,
             value<ComputationType> vct, value<ComputationResults> vcr, size_t y_block_cols,
             const value<model>& Mcurrent, const value<model>& Mbase,
652
653
654
655
656
657
             const locus_key& lk, const std::vector<double>& steps, double threshold)
        {
            if (positions.size()) {
                if (!locus_blocks.size()) {
                    _recompute(all_pops, lk, steps, effective_positions);
                }
658
                compute_along_interval<>(i0, cac, vct, vcr, y_block_cols, Mcurrent, Mbase, lk, chrom, effective_positions, locus_blocks);
659
660
661
662
663
664
                local_max = find_max(i0, vct, cac, threshold);
            } else {
                local_max.reset();
            }
        }

665
    std::pair<value<model_block_type>, value<model_block_type>>
666
667
668
669
670
        compute_at(const collection<population_value>& all_pops, double position)
        {
            if (locus_blocks.size() > 0) {
                auto it = std::find(effective_positions.begin(), effective_positions.end(), position);
                if (it != effective_positions.end()) {
671
672
                    auto ofs = it - effective_positions.begin();
                    return {locus_blocks[ofs], dominance_blocks[ofs]};
673
674
675
676
677
678
679
680
681
682
                }
            }
            local_max.reset();
            std::vector<double> tmp_pos = {position};
            positions.swap(tmp_pos);
            locus_key lk;
            if (mode == Joint) {
                lk = selection;
            }
            _recompute(all_pops, lk, positions, effective_positions);
683
684
685
            std::pair<value<model_block_type>, value<model_block_type>> ret;
            ret.first = locus_blocks[0];
            ret.second = dominance_blocks[0];
686
687
688
689
690
691
            locus_blocks.clear();
            effective_positions.clear();
            positions.swap(tmp_pos);
            return ret;
        }

692
693
694
695
696
697
698
699
700
    size_t
        count_positions() const
        {
            if (mode == Joint) {
                return positions.size() - selection->depth();
            }
            return positions.size();
        }

701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
    void
        _recompute(const collection<population_value>& all_pops, const locus_key& lk, const std::vector<double>& loci, std::vector<double>& effective_pos)
        {
            std::vector<double>::const_iterator i = loci.begin(), j = loci.end();
            effective_pos.clear();
            effective_pos.reserve(loci.size());
            if (lk) {
                for (; i != j; ++i) {
                    if (!lk->has(*i)) {
                        effective_pos.push_back(*i);
                    }
                }
            } else {
                effective_pos.assign(i, j);
            }
716
717
718
719
720
721
722
723
724
725
726
727
            /*MSG_DEBUG("search_interval recompute mode=" << (mode == Joint ? "Joint" : "Single") << " lk=" << lk);*/
            /* precompute locus probabilities first */
            MSG_DEBUG("Precomputing locus probabilities per individual");
            for (const auto& vpop: all_pops) {
                context_key ck(new context_key_struc(*vpop, chrom, loci));
                for (auto& x: make_collection<Disk>(locus_probabilities,
                                                    as_value(ck), as_value(lk), range<int>(0, (*vpop)->size(), 1),
                                                    as_value(ck->loci))) {
                    (void)*x;
                }
            }
            MSG_DEBUG("Computing parental origin probabilities");
728
729
730
731
732
733
734
            locus_blocks
                = compute_parental_origins_multipop(
                        all_pops,
                        as_value(chrom),
                        as_value(lk),
                        loci,
                        effective_pos);
735
            MSG_DEBUG("Computing dominance probabilities");
736
737
738
739
740
741
742
            dominance_blocks
                = compute_dominance_multipop(
                        all_pops,
                        as_value(chrom),
                        as_value(lk),
                        loci,
                        effective_pos);
743
        }
744
745

    test_result
746
        find_max(int i0, value<ComputationType> vct, computation_along_chromosome &cac, double threshold)
747
        {
748
            MSG_DEBUG("call to find_max");
749
750
            auto ct = *vct;
            auto last_computation =
751
752
753
754
                (ct == ComputationType::FTest    ? cac.ftest_pvalue
                :ct == ComputationType::FTestLOD ? cac.ftest_lod
                :ct == ComputationType::Chi2     ? cac.chi2
                :/* has to be Chi2LOD */           cac.chi2_lod).middleCols(i0, effective_positions.size());
755
            if (effective_positions.size() != (size_t)(last_computation.outerSize())) {
756
                MSG_ERROR("LOCI INCONSISTENT WITH COMPUTATION RESULT (" << effective_positions.size() << " vs " << last_computation.outerSize() << ")", "");
757
758
759
760
761
762
763
764
765
            }
            int i_max = -1;
            double max = -1;
            for (int i = 0; i < last_computation.outerSize(); ++i) {
                if (last_computation(0, i) >= max) {
                    max = last_computation(0, i);
                    i_max = i;
                }
            }
766
767
768
769
770
771
772
773
774
775

#ifdef SIGNAL_DISPLAY_ONELINER
        signal_display sd(last_computation.transpose(), i_max, max > threshold);
        MSG_DEBUG("[COMPUTATION] " << effective_positions.front() << sd << effective_positions.back() << " max=" << max << " at " << effective_positions[i_max]);
#else
        signal_display sd(*chrom, effective_positions, last_computation.transpose(), i_max, threshold);
        MSG_DEBUG("[COMPUTATION] " << effective_positions.front() << " ... " << effective_positions.back() << " max=" << max << " at " << effective_positions[i_max] << std::endl << sd);
#endif

        if (i_max == -1) {
776
777
778
                return {};
            }
            /*model_block_key k = locus_base_key;*/
779
780
781
782
783
784
            model_block_key mbk;
            if (mode == Joint) {
                mbk = model_block_key_struc::pop(chrom, selection + effective_positions[i_max]);
            } else {
                mbk = model_block_key_struc::pop(chrom, locus_key{} + effective_positions[i_max]);
            }
785
786
            /*model_block_key mbk = locus_base_key;*/
            /*mbk += std::make_pair(chrom, (*loci)[i_max]);*/
787
            /*MSG_DEBUG("locus_base_key " << locus_base_key << " mbk " << mbk);*/
788
789
790
791
792
793
/*
            MSG_DEBUG("last_computation@" << last_computation);
            MSG_QUEUE_FLUSH();
            MSG_DEBUG((*last_computation));
            MSG_QUEUE_FLUSH();
*/
794

795
796
            return {
                this, chrom,
797
798
                effective_positions[i_max], max, i_max, max > threshold,
                mbk,
799
800
801
                locus_blocks[i_max],
                dominance_blocks[i_max]->cols() ? model_block_key_struc::dominance(mbk) : model_block_key{},
                dominance_blocks[i_max]
802
            };
803
804
        }

805
806
807
808
809
810
    void
        select(const test_result& tr, value<model> M)
        {
            assert(chrom == tr.chrom);
            if (mode == Joint) {
                M->remove_block(model_block_key_struc::pop(chrom, selection));
811
                selection = tr.pop_block_key->loci;
812
            } else {
813
814
815
816
817
                selection = selection + tr.pop_block_key->loci;
            }
            MSG_DEBUG("Adding block " << tr.pop_block_key);
            M->add_block(tr.pop_block_key, tr.pop_block);
            if (tr.dom_block_key) {
818
819
820
                model ref = *M;
                ref.compute();
                M->add_block_if_test_is_good(tr.dom_block_key, tr.dom_block, ref);
821
            }
822
        }
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840

    void
        deselect(double position, const collection<population_value>& all_pops, value<model> M)
        {
            if (mode == Joint) {
                auto vmbk = model_block_key_struc::pop(chrom, selection);
                value<model_block_type> vblock = M->m_blocks[vmbk];
                M->remove_block(vmbk);
                /* Need to add the reduced block now */
                if (selection->depth() > 1) {
                    reduce(all_pops, position, vblock);
                    M->add_block(model_block_key_struc::pop(chrom, selection - position), vblock);
                }
            } else {
                locus_key lk;
                lk = lk + position;
                M->remove_block(model_block_key_struc::pop(chrom, lk));
            }
841
842
            locus_blocks.clear();
            selection = selection - position;
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
            /*deselect(position);*/
        }

    std::pair<model_block_key, model_block_key>
        select(double pos)
        {
            std::pair<model_block_key, model_block_key> ret;
            if (mode == Joint) {
                ret.second = model_block_key_struc::pop(chrom, selection);
            }
            selection = selection + pos;
            if (mode == Single) {
                ret.first = model_block_key_struc::pop(chrom, locus_key{} + pos);
            } else {
                ret.first = model_block_key_struc::pop(chrom, selection);
            }
            locus_blocks.clear();
            return ret;
        }

    std::pair<model_block_key, model_block_key>
        deselect(double pos)
        {
            std::pair<model_block_key, model_block_key> ret;
            if (mode == Joint) {
                ret.second = model_block_key_struc::pop(chrom, selection);
            } else {
                ret.second = model_block_key_struc::pop(chrom, locus_key{} + pos);
            }
            selection = selection - pos;
            if (mode == Joint && selection) {
                ret.first = model_block_key_struc::pop(chrom, selection);
            }
            locus_blocks.clear();
            return ret;
        }

    void
        reduce(const collection<population_value>& all_pops, double position, value<model_block_type>& vblock)
        {
883
            /* FIXME what about dominance?? */
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
            locus_key lk2 = selection - position;
            auto pop_it = all_pops.begin();
            auto pb = disassemble_parental_origins_multipop(chrom, selection->parent, *vblock, all_pops);
            std::vector<collection<parental_origin_per_locus_type>> all_popl;
            all_popl.reserve(pb.size());
            for (auto& vmat: pb) {
                const qtl_pop_type* pop = **pop_it++;
                context_key ck(new context_key_struc(pop, chrom, std::vector<double>()));
                MatrixXd red = selection->reduce(active_settings->parent_count_per_pop_per_chr
                                                    .find(pop->qtl_generation_name)->second.find(chrom)->second,
                                                 position);
                MatrixXd data = vmat->data * red;
                vmat->data = data;
                vmat->column_labels = get_stpom_data(ck, lk2->parent)->row_labels;
                all_popl.emplace_back();
                all_popl.back().push_back(vmat);
            }
            vblock = assemble_parental_origins_multipop(as_value(chrom),
                                                        as_value(lk2->parent),
                                                        all_popl,
904
905
                                                        all_pops,
                                                        true)[0];
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
        }
};


typedef std::map<chromosome_value, std::vector<search_interval_type>> search_interval_map_type;


inline
search_interval_map_type
skeleton_search_intervals()
{
    search_interval_map_type ret;

    for (const chromosome& chr: active_settings->map) {
        ret[&chr].emplace_back(Single, &chr);
921
922
    }

923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
    if (active_settings->skeleton_mode == "auto") {
        for (const chromosome& chr: active_settings->map) {
            double accept = -1.;
            for (double l: chr.condensed.marker_locus) {
                if (l > accept) {
                    ret[&chr].back().all_positions.push_back(l);
                    accept = l + active_settings->skeleton_interval;
                }
            }
        }
    } else if (active_settings->skeleton_mode == "manual") {
        for (const auto& name: active_settings->skeleton_markers) {
            bool found = false;
            for (const chromosome& chr: active_settings->map) {
                auto li = chr.condensed.marker_locus.begin();
                auto lj = chr.condensed.marker_locus.end();
                auto ni = chr.condensed.marker_name.begin();
                for (; li != lj; ++li, ++ni) {
                    if (*ni == name) {
                        ret[&chr].back().all_positions.push_back(*li);
                        found = true;
                        break;
                    }
                }
                if (found) {
                    break;
                }
            }
951
952
953
        }
    }

954
955
956
957
958
959
    return ret;
}


inline
search_interval_map_type
960
full_search_intervals(const std::map<chromosome_value, locus_key>& selection)
961
962
963
964
965
966
967
{
    search_interval_map_type ret;
    /* TODO: add definition of intervals to command line: chromosome, mode, beginning, end */

    for (const chromosome& chr: active_settings->map) {
        ret[&chr].emplace_back(Single, &chr);
        ret[&chr].back().all_positions = active_settings->estimation_loci[&chr];
968
969
970
971
972
        auto it = selection.find(&chr);
        if (it != selection.end()) {
            locus_key lk = std::make_shared<locus_key_struc>(active_settings->estimation_loci[&chr]);
            ret[&chr].back().selection = it->second & lk;
        }
973
974
975
976
    }

    return ret;
}
977
978


979
980
981
982
void
read_locus_list(std::string& s, settings_t* target);


983
struct model_manager {
984

985
    /* name of the studied single_trait */
986
987
988
989
990
    std::string trait_name;
    double threshold;
    /* populations used in this model */
    collection<population_value> all_pops;
    /* thy reference model */
991
    value<model> vMcurrent;
992
    size_t y_block_cols;
993
994
    /* thy maximum order for interaction blocks */
    size_t max_order;
995
996

    /* all steps, split by haplotypic intervals, inside which joint probabilities must be computed */
997
    search_interval_map_type search_intervals;
998
999
    /* the current selection split by haplotypic interval */
    std::map<chromosome_value, std::vector<locus_key>> selection;
1000
    /* a collection of intervals to not search in, if the user wishes so */