医学、机器学习等等,在统计结果时时长会用到这两个指标来说明数据的特性。
定义
敏感性:在金标准判断有病(阳性)人群中,检测出阳性的几率。真阳性。(检测出确实有病的能力)
特异性:在金标准判断无病(阴性)人群中,检测出阴性的几率。真阴性。(检测出确实没病的能力) :得到了阳性结果,但这个阳性结果是假的。即在金标准判断无病(阴性)人群中,检测出为阳性的几率。(没病,但却检测结果说有病),为误诊率。 假阴性率:得到了阴性结果,但这个阴性结果是假的。即在金标准判断有病(阳性)人群中,检测出为阴性的几率。(有病,但却检测结果说没病),为漏诊率。计算方法
Sensitivity and specificity:
True Positive (真正, TP)被模型预测为正的正样本;可以称作判断为真的正确率True Negative(真负 , TN)被模型预测为负的负样本 ;可以称作判断为假的正确率False Positive (假正, FP)被模型预测为正的负样本;可以称作误报率False Negative(假负 , FN)被模型预测为负的正样本;可以称作漏报率True Positive Rate(真正率 , TPR)或灵敏度(sensitivity) TPR = TP /(TP + FN) 正样本预测结果数 / 正样本实际数 True Negative Rate(真负率 , TNR)或特指度(specificity) TNR = TN /(TN + FP) 负样本预测结果数 / 负样本实际数 False Positive Rate (假正率, FPR) FPR = FP /(FP + TN) 被预测为正的负样本结果数 /负样本实际数False Negative Rate(假负率 , FNR) FNR = FN /(TP + FN) 被预测为负的正样本结果数 / 正样本实际数
假阳性率=假阳性人数÷金标准阴性人数
即: 假阳性率=b÷(b+d)
金标准 | 金标准 | |||
阳性(+) | 阴性(-) | 合计 | ||
某筛检方法 | 阳性(+) | a | b | a+b |
某筛检方法 | 阴性(-) | c | d | c+d |
合计 | a+c | b+d | N |
公式为:假阳性率=b/(b+d)×100%
(b:筛选为阳性,而标准分类为阴性的例数;d:阴性一致例数)
假阴性率=假阴性人数÷金标准阳性人数
即: β=c÷(a+c)
终于要用到这个玩意了,很激动,主要统计假阴假阳性率。
我的任务:
1. 评估Pacbio MHC variation calling 结果(CCS/non-CCS)与Hiseq数据结果的一致性。
2. 分别在不同深度梯度的区域完成以上评估,推断PB MHC做variation calling的最低深度。这里要将一个位点分为SNP、REF 和 LowQual,然后只去 SNP 和 REF 进行统计。
这是我一下午写出来的统计代码:
#!/usr/bin/env python# Author: LI ZHIXINimport sysimport pysamfrom collections import OrderedDictdef classify_DP(depth): if depth > 101: return 21 return ((depth-1)//5+1)def parse_rec(rec): sample = list(rec.samples)[0] # filter the Invalid line if not ('GQ' or 'GT' or 'DP') in rec.samples[sample].keys() or len(rec.alleles) <= 1: # continue return 1, "LowQual", rec.pos # filter the LowQual if rec.samples[sample]['GQ'] < 30: return rec.samples[sample]['DP'], "LowQual", rec.pos # filter the indel flag = 0 for one in rec.alleles: if len(one) != len(rec.ref): flag = 1 if flag == 1: return rec.samples[sample]['DP'], "LowQual", rec.pos if rec.samples[sample]['GT'] != (0, 0): # rec.qual > 30 # variation_dict[rec.pos] = ["snp", rec.alleles] return rec.samples[sample]['DP'], "snp", rec.pos elif rec.samples[sample]['GT'] == (0, 0): # variation_dict[rec.pos] = ["ref", rec.alleles] return rec.samples[sample]['DP'], "ref", rec.posdef read_gvcf(gvcf_file_path): variation_dict = OrderedDict() for i in range(1,22): variation_dict[i] = {} for j in ('LowQual', 'snp', 'ref'): variation_dict[i][j] = [] # pos_list = [] gvcf_file = pysam.VariantFile(gvcf_file_path) for rec in gvcf_file.fetch('chr6',28477796,33448354): DP, pos_type, pos = parse_rec(rec) if DP < 1 or DP > 20: continue # DP = classify_DP(DP) variation_dict[DP][pos_type].append(pos) # print(pos, DP, pos_type) gvcf_file.close() # return variation_dict, pos_list return variation_dictdef read_hiseq_gvcf(gvcf_file_path): variation_dict = OrderedDict() # for i in range(1,22): # variation_dict[i] = {} for j in ('LowQual', 'snp', 'ref'): variation_dict[j] = [] # pos_list = [] gvcf_file = pysam.VariantFile(gvcf_file_path) for rec in gvcf_file.fetch('chr6',28477796,33448354): DP, pos_type, pos = parse_rec(rec) DP = classify_DP(DP) variation_dict[pos_type].append(pos) # print(pos, DP, pos_type) gvcf_file.close() # return variation_dict, pos_list return variation_dictdef show_dict_diff_DP(Hiseq_unified_variation_dict, PB_non_CCS_variation_dict, outf, outf2): for DP in range(1,21): Hiseq_snp = set(Hiseq_unified_variation_dict['snp']) Hiseq_ref = set(Hiseq_unified_variation_dict['ref']) Hiseq_lowqual = set(Hiseq_unified_variation_dict['LowQual']) PB_snp = PB_non_CCS_variation_dict[DP]['snp'] PB_ref = PB_non_CCS_variation_dict[DP]['ref'] PB_lowqual = PB_non_CCS_variation_dict[DP]['LowQual'] total = set(PB_snp + PB_ref + PB_lowqual) Hiseq_snp = total & Hiseq_snp Hiseq_ref = total & Hiseq_ref Hiseq_lowqual = total & Hiseq_lowqual PB_snp = set(PB_snp) PB_ref = set(PB_ref) PB_lowqual = set(PB_lowqual) a = len(Hiseq_snp & PB_snp) b = len(Hiseq_ref & PB_snp) c = len(Hiseq_lowqual & PB_snp) d = len(Hiseq_snp & PB_ref) e = len(Hiseq_ref & PB_ref) f = len(Hiseq_lowqual & PB_ref) g = len(Hiseq_snp & PB_lowqual) h = len(Hiseq_ref & PB_lowqual) i = len(Hiseq_lowqual & PB_lowqual) Low_total = (g+h+i)/(a+b+c+d+e+f+g+h+i) if (a+b) == 0: PPV = "NA" else: PPV = a/(a+b) PPV = "%.4f"%(PPV) if (a+d) == 0: TPR = "NA" else: TPR = a/(a+d) TPR = "%.4f"%(TPR) print(str(DP)+" :\n", a,b,c,"\n",d,e,f,"\n",g,h,i,"\n", file=outf2, sep='\t', end='\n') print(DP, TPR, PPV, "%.4f"%Low_total, file=outf, sep='\t', end='\n')with open("./depth_stat.txt", "w") as outf: print("Depth", "TPR", "PPV", "Low_total", file=outf, sep='\t', end='\n') outf2 = open("raw.txt", "w") Hiseq_unified_variation_dict = read_hiseq_gvcf("./hiseq_call_gvcf/MHC_Hiseq.unified.gvcf.gz") PB_non_CCS_variation_dict = read_gvcf("./non_CCS_PB_call_gvcf/MHC_non_CCS.unified.gvcf.gz") show_dict_diff_DP(Hiseq_unified_variation_dict, PB_non_CCS_variation_dict, outf, outf2) outf2
又碰到一个高级python语法:在双层循环中如何退出外层循环? 我用了一个手动的flag,有其他好方法吗?
如何统计下机数据的覆盖度和深度?当然要比对之后才能统计,而且还要对比对做一些处理。
在计算一个位点是否是SNP、indel、Ref时,不仅要考虑ref、alts、qual、GQ,而且必须要把GT、DP考虑在内,所以说还是比较复杂的。
最后如何分析第二个问题,call variation的最低深度?
统计不同深度下的假阴假阳性率,看在什么深度下其达到饱和。