The impact of molecular size on similarity between very closely related molecules.

How molecule size affects the impact of single-atom changes on molecular similarity.

This post adapts the analysis I did previously of the impact of single-atom changes on calculated similarity to take into account molecule size. It’s kind of kind of a “part 2” of the post looking at the impact of molecular size on similarity

For the analysis I use the same sets (tranches) of molecules collected from PubChem in part 1. I randomly selected 10K molecules from each tranche. The code for determining the impact of single-atom changes in each molecule is the same as the earlier post.

The results look like this:

fp	type	tranche	shift	90%	95%
MFP2	bits	10	0.46(0.11)	0.32	0.29
MFP2	bits	20	0.64(0.09)	0.52	0.48
MFP2	bits	30	0.71(0.08)	0.60	0.57
MFP2	bits	40	0.77(0.07)	0.68	0.66
MFP2	bits	50	0.80(0.06)	0.72	0.69

As before, we can think about these values as thresholds for retrieving extremely similar molecules; here we define that to mean molecules that have the same connectivity and differ in the identity of a single atom.

When working with molecules with between 10 and 20 atoms and using the MFP2 fingerprint (Morgan fingerprint with radius 2) we need to use a Tanimoto similarity threshold of 0.48 to be 95% certain of retrieving extremely similar molecules. The threshold goes up to 0.66 when working with molecules with between 30 and 40 atoms. The average similarity between extremely similar molecules with between 10 and 20 atoms is 0.64 with a standard deviation of 0.09.

Unsurprisingly, the values increase with the size of the molecules in the tranche. This holds for all of the fingerprints.

Here’s the giant table of numbers:

fp	type	tranche	shift	90%	95%
MFP0	counts	10	0.77(0.07)	0.75	0.67
MFP0	counts	20	0.89(0.02)	0.87	0.86
MFP0	counts	30	0.92(0.01)	0.90	0.90
MFP0	counts	40	0.94(0.00)	0.94	0.94
MFP0	counts	50	0.96(0.00)	0.95	0.95
MFP1	counts	10	0.61(0.09)	0.50	0.45
MFP1	counts	20	0.78(0.04)	0.73	0.70
MFP1	counts	30	0.84(0.03)	0.80	0.78
MFP1	counts	40	0.88(0.02)	0.86	0.85
MFP1	counts	50	0.91(0.02)	0.89	0.88
MFP2	counts	10	0.49(0.10)	0.36	0.33
MFP2	counts	20	0.68(0.08)	0.58	0.55
MFP2	counts	30	0.75(0.06)	0.68	0.66
MFP2	counts	40	0.82(0.04)	0.76	0.75
MFP2	counts	50	0.85(0.04)	0.81	0.80
MFP3	counts	10	0.44(0.10)	0.32	0.29
MFP3	counts	20	0.60(0.10)	0.47	0.44
MFP3	counts	30	0.68(0.08)	0.58	0.55
MFP3	counts	40	0.75(0.07)	0.67	0.65
MFP3	counts	50	0.80(0.06)	0.73	0.71
MFP0	bits	10	0.77(0.10)	0.67	0.60
MFP0	bits	20	0.87(0.05)	0.80	0.78
MFP0	bits	30	0.89(0.04)	0.82	0.80
MFP0	bits	40	0.91(0.03)	0.86	0.83
MFP0	bits	50	0.91(0.04)	0.86	0.83
MFP1	bits	10	0.59(0.11)	0.47	0.41
MFP1	bits	20	0.75(0.07)	0.67	0.63
MFP1	bits	30	0.80(0.06)	0.73	0.71
MFP1	bits	40	0.84(0.05)	0.78	0.76
MFP1	bits	50	0.86(0.05)	0.80	0.78
MFP2	bits	10	0.46(0.11)	0.32	0.29
MFP2	bits	20	0.64(0.09)	0.52	0.48
MFP2	bits	30	0.71(0.08)	0.60	0.57
MFP2	bits	40	0.77(0.07)	0.68	0.66
MFP2	bits	50	0.80(0.06)	0.72	0.69
MFP3	bits	10	0.41(0.11)	0.29	0.26
MFP3	bits	20	0.55(0.11)	0.41	0.37
MFP3	bits	30	0.63(0.10)	0.50	0.46
MFP3	bits	40	0.70(0.09)	0.59	0.56
MFP3	bits	50	0.74(0.08)	0.64	0.60
FFP0	counts	10	0.91(0.13)	0.75	0.71
FFP0	counts	20	0.96(0.06)	0.88	0.86
FFP0	counts	30	0.98(0.04)	0.91	0.90
FFP0	counts	40	0.98(0.03)	0.94	0.94
FFP0	counts	50	0.99(0.02)	0.95	0.95
FFP1	counts	10	0.85(0.20)	0.57	0.50
FFP1	counts	20	0.93(0.11)	0.77	0.73
FFP1	counts	30	0.96(0.07)	0.84	0.82
FFP1	counts	40	0.97(0.05)	0.88	0.86
FFP1	counts	50	0.98(0.04)	0.91	0.89
FFP2	counts	10	0.81(0.25)	0.43	0.38
FFP2	counts	20	0.90(0.15)	0.65	0.61
FFP2	counts	30	0.94(0.11)	0.75	0.71
FFP2	counts	40	0.96(0.08)	0.81	0.79
FFP2	counts	50	0.97(0.06)	0.85	0.83
FFP3	counts	10	0.79(0.27)	0.37	0.33
FFP3	counts	20	0.88(0.19)	0.56	0.51
FFP3	counts	30	0.92(0.14)	0.67	0.62
FFP3	counts	40	0.94(0.11)	0.74	0.71
FFP3	counts	50	0.95(0.09)	0.79	0.76
FFP0	bits	10	0.92(0.16)	0.67	0.50
FFP0	bits	20	0.96(0.11)	0.80	0.71
FFP0	bits	30	0.97(0.08)	0.86	0.80
FFP0	bits	40	0.98(0.06)	1.00	0.83
FFP0	bits	50	0.99(0.05)	1.00	0.86
FFP1	bits	10	0.84(0.22)	0.50	0.43
FFP1	bits	20	0.92(0.14)	0.69	0.62
FFP1	bits	30	0.95(0.11)	0.77	0.71
FFP1	bits	40	0.96(0.08)	0.83	0.78
FFP1	bits	50	0.97(0.06)	0.88	0.82
FFP2	bits	10	0.78(0.28)	0.35	0.29
FFP2	bits	20	0.88(0.19)	0.56	0.50
FFP2	bits	30	0.92(0.14)	0.66	0.60
FFP2	bits	40	0.94(0.11)	0.74	0.69
FFP2	bits	50	0.95(0.09)	0.79	0.74
FFP3	bits	10	0.76(0.30)	0.30	0.25
FFP3	bits	20	0.85(0.23)	0.45	0.39
FFP3	bits	30	0.90(0.18)	0.57	0.51
FFP3	bits	40	0.92(0.14)	0.66	0.60
FFP3	bits	50	0.94(0.12)	0.72	0.66
RDK5	counts	10	0.36(0.19)	0.13	0.09
RDK5	counts	20	0.59(0.18)	0.35	0.29
RDK5	counts	30	0.69(0.14)	0.49	0.43
RDK5	counts	40	0.77(0.12)	0.60	0.55
RDK5	counts	50	0.81(0.10)	0.66	0.62
RDK7	counts	10	0.30(0.16)	0.11	0.08
RDK7	counts	20	0.53(0.20)	0.27	0.22
RDK7	counts	30	0.64(0.17)	0.40	0.35
RDK7	counts	40	0.73(0.14)	0.53	0.47
RDK7	counts	50	0.79(0.13)	0.60	0.55
RDK5	bits	10	0.36(0.18)	0.14	0.11
RDK5	bits	20	0.59(0.17)	0.36	0.30
RDK5	bits	30	0.68(0.15)	0.47	0.42
RDK5	bits	40	0.76(0.13)	0.58	0.52
RDK5	bits	50	0.79(0.12)	0.62	0.57
RDK7	bits	10	0.30(0.16)	0.12	0.09
RDK7	bits	20	0.55(0.19)	0.31	0.26
RDK7	bits	30	0.68(0.16)	0.46	0.41
RDK7	bits	40	0.79(0.13)	0.61	0.55
RDK7	bits	50	0.84(0.11)	0.68	0.63
TT	counts	10	0.38(0.24)	0.00	0.00
TT	counts	20	0.63(0.17)	0.41	0.35
TT	counts	30	0.72(0.13)	0.55	0.50
TT	counts	40	0.79(0.10)	0.66	0.62
TT	counts	50	0.84(0.08)	0.72	0.69
TT	bits	10	0.41(0.25)	0.08	0.00
TT	bits	20	0.64(0.17)	0.41	0.35
TT	bits	30	0.73(0.13)	0.55	0.50
TT	bits	40	0.79(0.10)	0.66	0.62
TT	bits	50	0.83(0.09)	0.71	0.68
AP	counts	10	0.61(0.09)	0.56	0.50
AP	counts	20	0.79(0.03)	0.75	0.73
AP	counts	30	0.85(0.02)	0.82	0.82
AP	counts	40	0.89(0.01)	0.88	0.88
AP	counts	50	0.91(0.01)	0.90	0.90
AP	bits	10	0.62(0.09)	0.56	0.50
AP	bits	20	0.79(0.04)	0.75	0.72
AP	bits	30	0.85(0.03)	0.82	0.81
AP	bits	40	0.89(0.02)	0.87	0.86
AP	bits	50	0.91(0.02)	0.89	0.88

import gzip
import glob
import pickle
import numpy as np

from rdkit import Chem
from rdkit import rdBase
import rdkit
print(rdkit.__version__)

import matplotlib.pyplot as plt
%matplotlib inline
plt.style.use('tableau-colorblind10')

2025.03.3

with gzip.open('./results/pubchem_tranches.pkl.gz','rb') as inf:
    mol_accum = pickle.load(inf)

We need to remove Hs from the input molecules, and we’ll only keep the first 10K from each tranche

import random
random.seed(0xa100f)
with rdBase.BlockLogs() as bl:
    for k in mol_accum:
        nv = []
        v = mol_accum[k]
        random.shuffle(v)
        v = v[:10000]
        for m in v:
            try:
                nm = Chem.RemoveHs(m)
            except Chem.MolSanitizeException:
                pass
            nv.append(nm)
        mol_accum[k] = nv

print([len(v) for v in mol_accum.values()])

[10000, 10000, 10000, 10000, 10000]

from rdkit.Chem import Draw
Draw.MolsToGridImage(mol_accum[20][:10])

Collect similarity values for bit-based and count-based fingerprints using

from rdkit import DataStructs
def get_modified_sims2(mol,maxRadius,fpgFunc,method):
    fpg = fpgFunc()
    fpFunc = getattr(fpg,method)
    basefp = fpFunc(mol)
    dm = Chem.GetDistanceMatrix(mol)
    res = []
    for aidx in range(mol.GetNumAtoms()):
        nm = Chem.Mol(mol)
        nm.GetAtomWithIdx(aidx).SetAtomicNum(0)
        count = sum(dm[aidx]<=maxRadius)
        res.append((count,DataStructs.TanimotoSimilarity(basefp,fpFunc(nm))))
    return res

try:
    import ipyparallel as ipp
    rc = ipp.Client()
    dview = rc[:]
    dview.execute('from rdkit import Chem')
    dview.execute('from rdkit.Chem import rdFingerprintGenerator')
    dview.execute('from rdkit import DataStructs')
except:
    print("could not use ipyparallel")
    dview = None

from collections import defaultdict
mfp_accum = defaultdict(lambda: defaultdict(lambda : defaultdict(list)))

for k in mol_accum:
    for radius in (0,1,2,3):
        print(f'tranche: {k}, radius {radius}')
        fpg = lambda r=radius:rdFingerprintGenerator.GetMorganGenerator(radius=r,fpSize=2048)
        lres = dview.map_sync(lambda x,r=radius,y=get_modified_sims2,
                              z=fpg:y(x,r,z,'GetCountFingerprint'),mol_accum[k])
        for entry in lres:
            mfp_accum[k]['counts'][radius].extend(entry)
        lres = dview.map_sync(lambda x,r=radius,y=get_modified_sims2,
                              z=fpg:y(x,r,z,'GetFingerprint'),mol_accum[k])
        for entry in lres:
            mfp_accum[k]['bits'][radius].extend(entry)

tranche: 10, radius 0
tranche: 10, radius 1
tranche: 10, radius 2
tranche: 10, radius 3
tranche: 20, radius 0
tranche: 20, radius 1
tranche: 20, radius 2
tranche: 20, radius 3
tranche: 30, radius 0
tranche: 30, radius 1
tranche: 30, radius 2
tranche: 30, radius 3
tranche: 40, radius 0
tranche: 40, radius 1
tranche: 40, radius 2
tranche: 40, radius 3
tranche: 50, radius 0
tranche: 50, radius 1
tranche: 50, radius 2
tranche: 50, radius 3

ffp_accum = defaultdict(lambda: defaultdict(lambda : defaultdict(list)))

for k in mol_accum:
    for radius in (0, 1,2,3):
        print(f'tranche: {k}, radius {radius}')
        fpg = lambda r=radius:rdFingerprintGenerator.GetMorganGenerator(radius=r,fpSize=2048,
                                                                        atomInvariantsGenerator=rdFingerprintGenerator.GetMorganFeatureAtomInvGen())
        lres = dview.map_sync(lambda x,r=radius,y=get_modified_sims2,
                              z=fpg:y(x,r,z,'GetCountFingerprint'),mol_accum[k])
        for entry in lres:
            ffp_accum[k]['counts'][radius].extend(entry)
        lres = dview.map_sync(lambda x,r=radius,y=get_modified_sims2,
                              z=fpg:y(x,r,z,'GetFingerprint'),mol_accum[k])
        for entry in lres:
            ffp_accum[k]['bits'][radius].extend(entry)

tranche: 10, radius 0
tranche: 10, radius 1
tranche: 10, radius 2
tranche: 10, radius 3
tranche: 20, radius 0
tranche: 20, radius 1
tranche: 20, radius 2
tranche: 20, radius 3
tranche: 30, radius 0
tranche: 30, radius 1
tranche: 30, radius 2
tranche: 30, radius 3
tranche: 40, radius 0
tranche: 40, radius 1
tranche: 40, radius 2
tranche: 40, radius 3
tranche: 50, radius 0
tranche: 50, radius 1
tranche: 50, radius 2
tranche: 50, radius 3

rdk_accum = defaultdict(lambda: defaultdict(lambda : defaultdict(list)))

for k in mol_accum:
    for radius in (5,7):
        print(f'tranche: {k}, radius {radius}')
        fpg = lambda r=radius:rdFingerprintGenerator.GetRDKitFPGenerator(maxPath=r,fpSize=2048)
        lres = dview.map_sync(lambda x,r=radius,y=get_modified_sims2,
                              z=fpg:y(x,r,z,'GetCountFingerprint'),mol_accum[k])
        for entry in lres:
            rdk_accum[k]['counts'][radius].extend(entry)
        lres = dview.map_sync(lambda x,r=radius,y=get_modified_sims2,
                              z=fpg:y(x,r,z,'GetFingerprint'),mol_accum[k])
        for entry in lres:
            rdk_accum[k]['bits'][radius].extend(entry)

tranche: 10, radius 5
tranche: 10, radius 7
tranche: 20, radius 5
tranche: 20, radius 7
tranche: 30, radius 5
tranche: 30, radius 7
tranche: 40, radius 5
tranche: 40, radius 7
tranche: 50, radius 5
tranche: 50, radius 7

tt_accum = defaultdict(lambda: defaultdict(lambda : defaultdict(list)))

for k in mol_accum:
    for radius in (3,):
        print(f'tranche: {k}, radius {radius}')
        fpg = lambda r=radius:rdFingerprintGenerator.GetTopologicalTorsionGenerator(fpSize=4096)
        lres = dview.map_sync(lambda x,r=radius,y=get_modified_sims2,
                              z=fpg:y(x,r,z,'GetCountFingerprint'),mol_accum[k])
        for entry in lres:
            tt_accum[k]['counts'][radius].extend(entry)
        lres = dview.map_sync(lambda x,r=radius,y=get_modified_sims2,
                              z=fpg:y(x,r,z,'GetFingerprint'),mol_accum[k])
        for entry in lres:
            tt_accum[k]['bits'][radius].extend(entry)

tranche: 10, radius 3
tranche: 20, radius 3
tranche: 30, radius 3
tranche: 40, radius 3
tranche: 50, radius 3

ap_accum = defaultdict(lambda: defaultdict(lambda : defaultdict(list)))

for k in mol_accum:
    for radius in (3,):
        print(f'tranche: {k}, radius {radius}')
        fpg = lambda r=radius:rdFingerprintGenerator.GetAtomPairGenerator(fpSize=4096)
        lres = dview.map_sync(lambda x,r=radius,y=get_modified_sims2,
                              z=fpg:y(x,r,z,'GetCountFingerprint'),mol_accum[k])
        for entry in lres:
            ap_accum[k]['counts'][radius].extend(entry)
        lres = dview.map_sync(lambda x,r=radius,y=get_modified_sims2,
                              z=fpg:y(x,r,z,'GetFingerprint'),mol_accum[k])
        for entry in lres:
            ap_accum[k]['bits'][radius].extend(entry)

tranche: 10, radius 3
tranche: 20, radius 3
tranche: 30, radius 3
tranche: 40, radius 3
tranche: 50, radius 3

import copy
def to_dict(dd):
    if not isinstance(dd,defaultdict):
        return copy.deepcopy(dd)
    res = {}
    for k,v in dd.items():
        res[k] = to_dict(v)
    return res

import pickle
import gzip
with gzip.open('./results/fp_perturbation.tranches.pkl.gz','wb+') as outf:
    pickle.dump((to_dict(mfp_accum),to_dict(ffp_accum),to_dict(rdk_accum),to_dict(ap_accum),to_dict(tt_accum)),outf)

!ls -lh ./results/fp_perturbation.tranches.pkl.gz

-rw-rw-r-- 1 glandrum glandrum 85M Jul  3 16:16 ./results/fp_perturbation.tranches.pkl.gz

print('''| fp | type | tranche| shift | 90% | 95% |
| -- | -- | -- | -- | -- | -- |''')

rs = (0, 1, 2, 3)
titl='MFP'
accum = mfp_accum

for which in ('counts','bits'):
    for rad in rs:
        for tranche in accum:
            tacc = accum[tranche]
            yps = [y for x,y in tacc[which][rad]]
            qs = np.quantile(yps,[0.1,0.05])
            print(f'| {titl}{rad} | {which} | {tranche} | {np.mean(yps):.2f}({np.std(yps):.2f}) | {qs[0]:.2f} | {qs[1]:.2f} |')
titl='FFP'
accum = ffp_accum

for which in ('counts','bits'):
    for rad in rs:
        for tranche in accum:
            tacc = accum[tranche]
            yps = [y for x,y in tacc[which][rad]]
            qs = np.quantile(yps,[0.1,0.05])
            print(f'| {titl}{rad} | {which} | {tranche} | {np.mean(yps):.2f}({np.std(yps):.2f}) | {qs[0]:.2f} | {qs[1]:.2f} |')

rs = (5,7)
titl='RDK'
accum = rdk_accum

for which in ('counts','bits'):
    for rad in rs:
        for tranche in accum:
            tacc = accum[tranche]
            yps = [y for x,y in tacc[which][rad]]
            qs = np.quantile(yps,[0.1,0.05])
            print(f'| {titl}{rad} | {which} | {tranche} | {np.mean(yps):.2f}({np.std(yps):.2f}) | {qs[0]:.2f} | {qs[1]:.2f} |')


rs = (3, )
titl='TT'
accum = tt_accum

for which in ('counts','bits'):
    for rad in rs:
        for tranche in accum:
            tacc = accum[tranche]
            yps = [y for x,y in tacc[which][rad]]
            qs = np.quantile(yps,[0.1,0.05])
            print(f'| {titl} | {which} | {tranche} | {np.mean(yps):.2f}({np.std(yps):.2f}) | {qs[0]:.2f} | {qs[1]:.2f} |')
            
rs = (3, )
titl='AP'
accum = ap_accum

for which in ('counts','bits'):
    for rad in rs:
        for tranche in accum:
            tacc = accum[tranche]
            yps = [y for x,y in tacc[which][rad]]
            qs = np.quantile(yps,[0.1,0.05])
            print(f'| {titl} | {which} | {tranche} | {np.mean(yps):.2f}({np.std(yps):.2f}) | {qs[0]:.2f} | {qs[1]:.2f} |')
         

| fp | type | tranche| shift | 90% | 95% |
| -- | -- | -- | -- | -- | -- |
| MFP0 | counts | 10 | 0.77(0.07) | 0.75 | 0.67 |
| MFP0 | counts | 20 | 0.89(0.02) | 0.87 | 0.86 |
| MFP0 | counts | 30 | 0.92(0.01) | 0.90 | 0.90 |
| MFP0 | counts | 40 | 0.94(0.00) | 0.94 | 0.94 |
| MFP0 | counts | 50 | 0.96(0.00) | 0.95 | 0.95 |
| MFP1 | counts | 10 | 0.61(0.09) | 0.50 | 0.45 |
| MFP1 | counts | 20 | 0.78(0.04) | 0.73 | 0.70 |
| MFP1 | counts | 30 | 0.84(0.03) | 0.80 | 0.78 |
| MFP1 | counts | 40 | 0.88(0.02) | 0.86 | 0.85 |
| MFP1 | counts | 50 | 0.91(0.02) | 0.89 | 0.88 |
| MFP2 | counts | 10 | 0.49(0.10) | 0.36 | 0.33 |
| MFP2 | counts | 20 | 0.68(0.08) | 0.58 | 0.55 |
| MFP2 | counts | 30 | 0.75(0.06) | 0.68 | 0.66 |
| MFP2 | counts | 40 | 0.82(0.04) | 0.76 | 0.75 |
| MFP2 | counts | 50 | 0.85(0.04) | 0.81 | 0.80 |
| MFP3 | counts | 10 | 0.44(0.10) | 0.32 | 0.29 |
| MFP3 | counts | 20 | 0.60(0.10) | 0.47 | 0.44 |
| MFP3 | counts | 30 | 0.68(0.08) | 0.58 | 0.55 |
| MFP3 | counts | 40 | 0.75(0.07) | 0.67 | 0.65 |
| MFP3 | counts | 50 | 0.80(0.06) | 0.73 | 0.71 |
| MFP0 | bits | 10 | 0.77(0.10) | 0.67 | 0.60 |
| MFP0 | bits | 20 | 0.87(0.05) | 0.80 | 0.78 |
| MFP0 | bits | 30 | 0.89(0.04) | 0.82 | 0.80 |
| MFP0 | bits | 40 | 0.91(0.03) | 0.86 | 0.83 |
| MFP0 | bits | 50 | 0.91(0.04) | 0.86 | 0.83 |
| MFP1 | bits | 10 | 0.59(0.11) | 0.47 | 0.41 |
| MFP1 | bits | 20 | 0.75(0.07) | 0.67 | 0.63 |
| MFP1 | bits | 30 | 0.80(0.06) | 0.73 | 0.71 |
| MFP1 | bits | 40 | 0.84(0.05) | 0.78 | 0.76 |
| MFP1 | bits | 50 | 0.86(0.05) | 0.80 | 0.78 |
| MFP2 | bits | 10 | 0.46(0.11) | 0.32 | 0.29 |
| MFP2 | bits | 20 | 0.64(0.09) | 0.52 | 0.48 |
| MFP2 | bits | 30 | 0.71(0.08) | 0.60 | 0.57 |
| MFP2 | bits | 40 | 0.77(0.07) | 0.68 | 0.66 |
| MFP2 | bits | 50 | 0.80(0.06) | 0.72 | 0.69 |
| MFP3 | bits | 10 | 0.41(0.11) | 0.29 | 0.26 |
| MFP3 | bits | 20 | 0.55(0.11) | 0.41 | 0.37 |
| MFP3 | bits | 30 | 0.63(0.10) | 0.50 | 0.46 |
| MFP3 | bits | 40 | 0.70(0.09) | 0.59 | 0.56 |
| MFP3 | bits | 50 | 0.74(0.08) | 0.64 | 0.60 |
| FFP0 | counts | 10 | 0.91(0.13) | 0.75 | 0.71 |
| FFP0 | counts | 20 | 0.96(0.06) | 0.88 | 0.86 |
| FFP0 | counts | 30 | 0.98(0.04) | 0.91 | 0.90 |
| FFP0 | counts | 40 | 0.98(0.03) | 0.94 | 0.94 |
| FFP0 | counts | 50 | 0.99(0.02) | 0.95 | 0.95 |
| FFP1 | counts | 10 | 0.85(0.20) | 0.57 | 0.50 |
| FFP1 | counts | 20 | 0.93(0.11) | 0.77 | 0.73 |
| FFP1 | counts | 30 | 0.96(0.07) | 0.84 | 0.82 |
| FFP1 | counts | 40 | 0.97(0.05) | 0.88 | 0.86 |
| FFP1 | counts | 50 | 0.98(0.04) | 0.91 | 0.89 |
| FFP2 | counts | 10 | 0.81(0.25) | 0.43 | 0.38 |
| FFP2 | counts | 20 | 0.90(0.15) | 0.65 | 0.61 |
| FFP2 | counts | 30 | 0.94(0.11) | 0.75 | 0.71 |
| FFP2 | counts | 40 | 0.96(0.08) | 0.81 | 0.79 |
| FFP2 | counts | 50 | 0.97(0.06) | 0.85 | 0.83 |
| FFP3 | counts | 10 | 0.79(0.27) | 0.37 | 0.33 |
| FFP3 | counts | 20 | 0.88(0.19) | 0.56 | 0.51 |
| FFP3 | counts | 30 | 0.92(0.14) | 0.67 | 0.62 |
| FFP3 | counts | 40 | 0.94(0.11) | 0.74 | 0.71 |
| FFP3 | counts | 50 | 0.95(0.09) | 0.79 | 0.76 |
| FFP0 | bits | 10 | 0.92(0.16) | 0.67 | 0.50 |
| FFP0 | bits | 20 | 0.96(0.11) | 0.80 | 0.71 |
| FFP0 | bits | 30 | 0.97(0.08) | 0.86 | 0.80 |
| FFP0 | bits | 40 | 0.98(0.06) | 1.00 | 0.83 |
| FFP0 | bits | 50 | 0.99(0.05) | 1.00 | 0.86 |
| FFP1 | bits | 10 | 0.84(0.22) | 0.50 | 0.43 |
| FFP1 | bits | 20 | 0.92(0.14) | 0.69 | 0.62 |
| FFP1 | bits | 30 | 0.95(0.11) | 0.77 | 0.71 |
| FFP1 | bits | 40 | 0.96(0.08) | 0.83 | 0.78 |
| FFP1 | bits | 50 | 0.97(0.06) | 0.88 | 0.82 |
| FFP2 | bits | 10 | 0.78(0.28) | 0.35 | 0.29 |
| FFP2 | bits | 20 | 0.88(0.19) | 0.56 | 0.50 |
| FFP2 | bits | 30 | 0.92(0.14) | 0.66 | 0.60 |
| FFP2 | bits | 40 | 0.94(0.11) | 0.74 | 0.69 |
| FFP2 | bits | 50 | 0.95(0.09) | 0.79 | 0.74 |
| FFP3 | bits | 10 | 0.76(0.30) | 0.30 | 0.25 |
| FFP3 | bits | 20 | 0.85(0.23) | 0.45 | 0.39 |
| FFP3 | bits | 30 | 0.90(0.18) | 0.57 | 0.51 |
| FFP3 | bits | 40 | 0.92(0.14) | 0.66 | 0.60 |
| FFP3 | bits | 50 | 0.94(0.12) | 0.72 | 0.66 |
| RDK5 | counts | 10 | 0.36(0.19) | 0.13 | 0.09 |
| RDK5 | counts | 20 | 0.59(0.18) | 0.35 | 0.29 |
| RDK5 | counts | 30 | 0.69(0.14) | 0.49 | 0.43 |
| RDK5 | counts | 40 | 0.77(0.12) | 0.60 | 0.55 |
| RDK5 | counts | 50 | 0.81(0.10) | 0.66 | 0.62 |
| RDK7 | counts | 10 | 0.30(0.16) | 0.11 | 0.08 |
| RDK7 | counts | 20 | 0.53(0.20) | 0.27 | 0.22 |
| RDK7 | counts | 30 | 0.64(0.17) | 0.40 | 0.35 |
| RDK7 | counts | 40 | 0.73(0.14) | 0.53 | 0.47 |
| RDK7 | counts | 50 | 0.79(0.13) | 0.60 | 0.55 |
| RDK5 | bits | 10 | 0.36(0.18) | 0.14 | 0.11 |
| RDK5 | bits | 20 | 0.59(0.17) | 0.36 | 0.30 |
| RDK5 | bits | 30 | 0.68(0.15) | 0.47 | 0.42 |
| RDK5 | bits | 40 | 0.76(0.13) | 0.58 | 0.52 |
| RDK5 | bits | 50 | 0.79(0.12) | 0.62 | 0.57 |
| RDK7 | bits | 10 | 0.30(0.16) | 0.12 | 0.09 |
| RDK7 | bits | 20 | 0.55(0.19) | 0.31 | 0.26 |
| RDK7 | bits | 30 | 0.68(0.16) | 0.46 | 0.41 |
| RDK7 | bits | 40 | 0.79(0.13) | 0.61 | 0.55 |
| RDK7 | bits | 50 | 0.84(0.11) | 0.68 | 0.63 |
| TT | counts | 10 | 0.38(0.24) | 0.00 | 0.00 |
| TT | counts | 20 | 0.63(0.17) | 0.41 | 0.35 |
| TT | counts | 30 | 0.72(0.13) | 0.55 | 0.50 |
| TT | counts | 40 | 0.79(0.10) | 0.66 | 0.62 |
| TT | counts | 50 | 0.84(0.08) | 0.72 | 0.69 |
| TT | bits | 10 | 0.41(0.25) | 0.08 | 0.00 |
| TT | bits | 20 | 0.64(0.17) | 0.41 | 0.35 |
| TT | bits | 30 | 0.73(0.13) | 0.55 | 0.50 |
| TT | bits | 40 | 0.79(0.10) | 0.66 | 0.62 |
| TT | bits | 50 | 0.83(0.09) | 0.71 | 0.68 |
| AP | counts | 10 | 0.61(0.09) | 0.56 | 0.50 |
| AP | counts | 20 | 0.79(0.03) | 0.75 | 0.73 |
| AP | counts | 30 | 0.85(0.02) | 0.82 | 0.82 |
| AP | counts | 40 | 0.89(0.01) | 0.88 | 0.88 |
| AP | counts | 50 | 0.91(0.01) | 0.90 | 0.90 |
| AP | bits | 10 | 0.62(0.09) | 0.56 | 0.50 |
| AP | bits | 20 | 0.79(0.04) | 0.75 | 0.72 |
| AP | bits | 30 | 0.85(0.03) | 0.82 | 0.81 |
| AP | bits | 40 | 0.89(0.02) | 0.87 | 0.86 |
| AP | bits | 50 | 0.91(0.02) | 0.89 | 0.88 |