Archie #22
Archie #22
Conversation
| #I think these parameters are NOT necessary for ArchIE - just retained for the signature of preprocess.process_data | ||
| match_bonus = 1 | ||
| max_mismatch = 1 | ||
| mismatch_penalty = 1 |
There was a problem hiding this comment.
These parameters are necessary for calculating the S* scores, which are also input features in the logistic regression.
match_bonus = 5000
max_mismatch = 5
mismatch_penalty = -10000
| mut_num, hap_num = tgt_gt.shape | ||
| iv = np.ones((hap_num, 1)) | ||
| counts = tgt_gt*np.matmul(tgt_gt, iv) | ||
| spectra = np.array([np.bincount(np.array(counts[:,idx] > 0).astype('int8'), minlength=hap_num+1) for idx in range(hap_num)]) |
There was a problem hiding this comment.
Line 70 is not correct. The spectra here only contain "single-ton", because counts[:,idx] > 0 returns a boolean array and FALSEs and TRUEs are regarded as 0s and 1s.
There was a problem hiding this comment.
Therefore, what this line founds is the total number of mutations that a haplotype contains.
There was a problem hiding this comment.
The negative number problem was caused by astype('int8'), because the range of int8 is from -128 to 127, when simulating large sample sizes (e.g. 100 diploid target individuals), negative numbers may occur. To fix it, we can change astype('int8') to astype('int64')
| #reading of data, preprocessing - i.e., calculating statistics -, and obtaining & labeling of true tracts | ||
| for replicate1, folder in enumerate(os.listdir(output_dir)): |
There was a problem hiding this comment.
Using enumerate makes the order of replicates in the feature table not consistent with the order of the folders containing simulated data
There was a problem hiding this comment.
This makes it difficult to check the correctness of the calculation for different features.
| dist_skew = sps.skew(tgt_dist, axis=1) | ||
| dist_kurtosis = sps.kurtosis(tgt_dist, axis=1) |
There was a problem hiding this comment.
skew and kurtosis may be np.nan occasionally, we could replace nan with 0
No description provided.