Filtering individual datasets in a database
A user of a genetic database may create and build upon their family tree in the database. For new users, creating a family tree can be difficult and time consuming. Even for users with established family trees, extending their family tree is a challenge requiring extensive research. Disclosed herein are embodiments for assisting users of a genetic database with building their family trees. In some embodiments, a method for assisting with constructing family trees includes receiving a target individual's genetic dataset. The method identifies a plurality of matched individuals who genetically match the target individual. The method identifies potential ancestors who are potential common ancestors between the target individual and one of the matched individuals. The method inputs a set of features related to the target individual to a machine learning model and filters the potential common ancestors to determine a subset of likely common ancestors for the target individual.
1. A computer-implemented method, comprising:
receiving a target individual genetic dataset associated with a target individual;
identifying a plurality of matched individuals who genetically match with the target individual;
identifying a plurality of potential ancestors who are potential common ancestors between the target individual and one of the matched individuals;
inputting a set of features related to the target individual to a machine learning model, wherein training of the machine learning model comprises:
generating a plurality of training samples, wherein generating one of the plurality of training samples comprises:
identifying a training target individual, the training target individual having a genetic dataset,
identifying ancestors of the training target individual from one or more family trees of the training target individual,
determining, based on existing family trees that includes the training target individuals, whether each of the ancestors is a direct-line ancestor of the training target individual, and
assigning a positive label to a particular ancestor responsive to the particular ancestor being a direct-line ancestor of the training target individual; and
training the machine learning model using the plurality of training samples;
filtering the plurality of potential common ancestors using the machine learning model to identify a subset of the potential common ancestors of the target individual.
2. The computer-implemented method of claim 1 , wherein the target individual and the plurality of matched individuals are related by identity by descent (IBD).
3. The computer-implemented method of claim 1 , wherein a potential common ancestor of the plurality of potential common ancestors is documented in a family tree of one of the matched individuals.
4. The computer-implemented method of claim 1 , wherein identifying a matched individual of the plurality of matched individuals comprises:
receiving a candidate individual genetic dataset of a candidate individual;
identifying matched genetic segments between the candidate individual genetic dataset and the target individual;
measuring a total length of the matched genetic segments in centimorgans; and
classifying the candidate individual as the matched individual responsive to the total length exceeding a threshold.
5. The computer-implemented method of claim 4 , wherein the set of features includes the total length of the matched genetic segments in centimorgans.
6. The computer-implemented method of claim 1 , wherein the set of features includes a generation difference between a potential common ancestor and a matched individual.
7. The computer-implemented method of claim 1 , wherein the set of features includes an age difference between the target individual and a potential common ancestor.
8. The computer-implemented method of claim 1 , wherein the set of features includes a percentage of descendants in a family tree of a potential common ancestor who are matched individuals of the target individual.
9. The computer-implemented method of claim 1 , wherein the machine learning model is a supervised learning model.
10. The computer-implemented method of claim 1 , wherein the potential common ancestors in the subset of the potential common ancestors are more likely to be direct-line ancestors of the target individual than other potential common ancestors that are filtered out by the machine learning model.
11. The computer-implemented method of claim 1 , wherein the machine learning model is configured to provide at least a label that a potential common ancestor is a direct-line ancestor or not.
12. The computer-implemented method of claim 1 , wherein training of the machine learning model comprises:
generating a plurality of training samples, the plurality of training samples comprising a set of positive training samples and a set of negative training samples;
extracting training features of the plurality of training samples to generate a plurality of feature vectors, each feature vector corresponding to one of the training samples;
inputting the plurality of feature vectors to the machine learning model;
using the machine learning model to predict labels of one or more ancestors in the training samples;
determining an objective function that compares the predicted labels to actual labels of the training samples; and
adjusting one or more weights of the machine learning model based on the objective function.
13. A non-transitory computer readable medium configured to store computer code comprising instructions, the instructions, when executed by one or more processors, cause the one or more processors to perform steps comprising:
receiving a target individual genetic dataset associated with a target individual;
identifying a plurality of matched individuals who genetically match with the target individual;
identifying a plurality of potential ancestors who are potential common ancestors between the target individual and one of the matched individuals;
inputting a set of features related to the target individual to a machine learning model, wherein training of the machine learning model comprises:
generating a plurality of training samples, wherein generating one of the plurality of training samples comprises:
identifying a training target individual, the training target individual having a genetic dataset,
identifying ancestors of the training target individual from one or more family trees of the training target individual,
determining, based on existing family trees that includes the training target individuals, whether each of the ancestors is a direct-line ancestor of the training target individual, and
assigning a positive label to a particular ancestor responsive to the particular ancestor being a direct-line ancestor of the training target individual; and
training the machine learning model using the plurality of training samples; and
filtering the plurality of potential common ancestors using the machine learning model to identify a subset of the potential common ancestors of the target individual.
14. The non-transitory computer readable medium of claim 13 , wherein a potential common ancestor of the plurality of potential common ancestors is documented in a family tree of one of the matched individuals.
15. The non-transitory computer readable medium of claim 13 , wherein identifying a matched individual of the plurality of matched individuals comprises:
receiving a candidate individual genetic dataset of a candidate individual;
identifying matched genetic segments between the candidate individual genetic dataset and the target individual;
measuring a total length of the matched genetic segments in centimorgans; and
classifying the candidate individual as the matched individual responsive to the total length exceeding a threshold.
16. The non-transitory computer readable medium of claim 13 , wherein the machine learning model is a supervised learning model.
17. The non-transitory computer readable medium of claim 13 , wherein training of the machine learning model comprises:
generating a plurality of training samples, the plurality of training samples comprising a set of positive training samples and a set of negative training samples;
extracting training features of the plurality of training samples to generate a plurality of feature vectors, each feature vector corresponding to one of the training samples;
inputting the plurality of feature vectors to the machine learning model;
using the machine learning model to predict labels of one or more ancestors in the training samples;
determining an objective function that compares the predicted labels to actual labels of the training samples; and
adjusting one or more weights of the machine learning model based on the objective function.
18. A system comprising:
one or more processors; and
memory configured to store computer code comprising instructions, the instructions, when executed by one or more processors, cause the one or more processors to perform steps comprising:
receiving a target individual genetic dataset associated with a target individual;
identifying a plurality of matched individuals who genetically match with the target individual;
identifying a plurality of potential ancestors who are potential common ancestors between the target individual and one of the matched individuals;
inputting a set of features related to the target individual to a machine learning model, wherein training of the machine learning model comprises:
generating a plurality of training samples, wherein generating one of the plurality of training samples comprises:
identifying a training target individual, the training target individual having a genetic dataset,
identifying ancestors of the training target individual from one or more family trees of the training target individual,
determining, based on existing family trees that includes the training target individuals, whether each of the ancestors is a direct-line ancestor of the training target individual, and
assigning a positive label to a particular ancestor responsive to the particular ancestor being a direct-line ancestor of the training target individual, and training the machine learning model using the plurality of training samples; and
filtering the plurality of potential common ancestors using the machine learning model to identify a subset of the potential common ancestors of the target individual.